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 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/export.h>
30 #include <linux/types.h>
31 #include <linux/times.h>
32 #include <linux/socket.h>
33 #include <linux/sockios.h>
34 #include <linux/net.h>
35 #include <linux/route.h>
36 #include <linux/netdevice.h>
37 #include <linux/in6.h>
38 #include <linux/mroute6.h>
39 #include <linux/init.h>
40 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <linux/nsproxy.h>
44 #include <linux/slab.h>
45 #include <net/net_namespace.h>
48 #include <net/ip6_fib.h>
49 #include <net/ip6_route.h>
50 #include <net/ndisc.h>
51 #include <net/addrconf.h>
53 #include <linux/rtnetlink.h>
56 #include <net/netevent.h>
57 #include <net/netlink.h>
59 #include <asm/uaccess.h>
62 #include <linux/sysctl.h>
65 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
66 const struct in6_addr *dest);
67 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
68 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
69 static unsigned int ip6_mtu(const struct dst_entry *dst);
70 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
71 static void ip6_dst_destroy(struct dst_entry *);
72 static void ip6_dst_ifdown(struct dst_entry *,
73 struct net_device *dev, int how);
74 static int ip6_dst_gc(struct dst_ops *ops);
76 static int ip6_pkt_discard(struct sk_buff *skb);
77 static int ip6_pkt_discard_out(struct sk_buff *skb);
78 static void ip6_link_failure(struct sk_buff *skb);
79 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
81 #ifdef CONFIG_IPV6_ROUTE_INFO
82 static struct rt6_info *rt6_add_route_info(struct net *net,
83 const struct in6_addr *prefix, int prefixlen,
84 const struct in6_addr *gwaddr, int ifindex,
86 static struct rt6_info *rt6_get_route_info(struct net *net,
87 const struct in6_addr *prefix, int prefixlen,
88 const struct in6_addr *gwaddr, int ifindex);
91 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
93 struct rt6_info *rt = (struct rt6_info *) dst;
94 struct inet_peer *peer;
97 if (!(rt->dst.flags & DST_HOST))
101 rt6_bind_peer(rt, 1);
103 peer = rt->rt6i_peer;
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, const void *daddr)
126 struct in6_addr *p = &rt->rt6i_gateway;
128 if (!ipv6_addr_any(p))
129 return (const void *) p;
133 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst, const void *daddr)
135 struct rt6_info *rt = (struct rt6_info *) dst;
138 daddr = choose_neigh_daddr(rt, daddr);
139 n = __ipv6_neigh_lookup(&nd_tbl, dst->dev, daddr);
142 return neigh_create(&nd_tbl, daddr, dst->dev);
145 static int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
147 struct neighbour *n = __ipv6_neigh_lookup(&nd_tbl, dev, &rt->rt6i_gateway);
149 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
153 dst_set_neighbour(&rt->dst, n);
158 static struct dst_ops ip6_dst_ops_template = {
160 .protocol = cpu_to_be16(ETH_P_IPV6),
163 .check = ip6_dst_check,
164 .default_advmss = ip6_default_advmss,
166 .cow_metrics = ipv6_cow_metrics,
167 .destroy = ip6_dst_destroy,
168 .ifdown = ip6_dst_ifdown,
169 .negative_advice = ip6_negative_advice,
170 .link_failure = ip6_link_failure,
171 .update_pmtu = ip6_rt_update_pmtu,
172 .local_out = __ip6_local_out,
173 .neigh_lookup = ip6_neigh_lookup,
176 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
178 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
180 return mtu ? : dst->dev->mtu;
183 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
187 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
193 static struct dst_ops ip6_dst_blackhole_ops = {
195 .protocol = cpu_to_be16(ETH_P_IPV6),
196 .destroy = ip6_dst_destroy,
197 .check = ip6_dst_check,
198 .mtu = ip6_blackhole_mtu,
199 .default_advmss = ip6_default_advmss,
200 .update_pmtu = ip6_rt_blackhole_update_pmtu,
201 .cow_metrics = ip6_rt_blackhole_cow_metrics,
202 .neigh_lookup = ip6_neigh_lookup,
205 static const u32 ip6_template_metrics[RTAX_MAX] = {
206 [RTAX_HOPLIMIT - 1] = 255,
209 static struct rt6_info ip6_null_entry_template = {
211 .__refcnt = ATOMIC_INIT(1),
214 .error = -ENETUNREACH,
215 .input = ip6_pkt_discard,
216 .output = ip6_pkt_discard_out,
218 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
219 .rt6i_protocol = RTPROT_KERNEL,
220 .rt6i_metric = ~(u32) 0,
221 .rt6i_ref = ATOMIC_INIT(1),
224 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
226 static int ip6_pkt_prohibit(struct sk_buff *skb);
227 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
229 static struct rt6_info ip6_prohibit_entry_template = {
231 .__refcnt = ATOMIC_INIT(1),
235 .input = ip6_pkt_prohibit,
236 .output = ip6_pkt_prohibit_out,
238 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
239 .rt6i_protocol = RTPROT_KERNEL,
240 .rt6i_metric = ~(u32) 0,
241 .rt6i_ref = ATOMIC_INIT(1),
244 static struct rt6_info ip6_blk_hole_entry_template = {
246 .__refcnt = ATOMIC_INIT(1),
250 .input = dst_discard,
251 .output = dst_discard,
253 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
254 .rt6i_protocol = RTPROT_KERNEL,
255 .rt6i_metric = ~(u32) 0,
256 .rt6i_ref = ATOMIC_INIT(1),
261 /* allocate dst with ip6_dst_ops */
262 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops,
263 struct net_device *dev,
266 struct rt6_info *rt = dst_alloc(ops, dev, 0, 0, flags);
269 memset(&rt->rt6i_table, 0,
270 sizeof(*rt) - sizeof(struct dst_entry));
275 static void ip6_dst_destroy(struct dst_entry *dst)
277 struct rt6_info *rt = (struct rt6_info *)dst;
278 struct inet6_dev *idev = rt->rt6i_idev;
279 struct inet_peer *peer = rt->rt6i_peer;
281 if (!(rt->dst.flags & DST_HOST))
282 dst_destroy_metrics_generic(dst);
285 rt->rt6i_idev = NULL;
289 rt->rt6i_peer = NULL;
294 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
296 static u32 rt6_peer_genid(void)
298 return atomic_read(&__rt6_peer_genid);
301 void rt6_bind_peer(struct rt6_info *rt, int create)
303 struct inet_peer *peer;
305 peer = inet_getpeer_v6(&rt->rt6i_dst.addr, create);
306 if (peer && cmpxchg(&rt->rt6i_peer, NULL, peer) != NULL)
309 rt->rt6i_peer_genid = rt6_peer_genid();
312 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
315 struct rt6_info *rt = (struct rt6_info *)dst;
316 struct inet6_dev *idev = rt->rt6i_idev;
317 struct net_device *loopback_dev =
318 dev_net(dev)->loopback_dev;
320 if (dev != loopback_dev && idev && idev->dev == dev) {
321 struct inet6_dev *loopback_idev =
322 in6_dev_get(loopback_dev);
324 rt->rt6i_idev = loopback_idev;
330 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
332 return (rt->rt6i_flags & RTF_EXPIRES) &&
333 time_after(jiffies, rt->dst.expires);
336 static inline int rt6_need_strict(const struct in6_addr *daddr)
338 return ipv6_addr_type(daddr) &
339 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
343 * Route lookup. Any table->tb6_lock is implied.
346 static inline struct rt6_info *rt6_device_match(struct net *net,
348 const struct in6_addr *saddr,
352 struct rt6_info *local = NULL;
353 struct rt6_info *sprt;
355 if (!oif && ipv6_addr_any(saddr))
358 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
359 struct net_device *dev = sprt->dst.dev;
362 if (dev->ifindex == oif)
364 if (dev->flags & IFF_LOOPBACK) {
365 if (!sprt->rt6i_idev ||
366 sprt->rt6i_idev->dev->ifindex != oif) {
367 if (flags & RT6_LOOKUP_F_IFACE && oif)
369 if (local && (!oif ||
370 local->rt6i_idev->dev->ifindex == oif))
376 if (ipv6_chk_addr(net, saddr, dev,
377 flags & RT6_LOOKUP_F_IFACE))
386 if (flags & RT6_LOOKUP_F_IFACE)
387 return net->ipv6.ip6_null_entry;
393 #ifdef CONFIG_IPV6_ROUTER_PREF
394 static void rt6_probe(struct rt6_info *rt)
396 struct neighbour *neigh;
398 * Okay, this does not seem to be appropriate
399 * for now, however, we need to check if it
400 * is really so; aka Router Reachability Probing.
402 * Router Reachability Probe MUST be rate-limited
403 * to no more than one per minute.
406 neigh = rt ? dst_get_neighbour_noref(&rt->dst) : NULL;
407 if (!neigh || (neigh->nud_state & NUD_VALID))
409 read_lock_bh(&neigh->lock);
410 if (!(neigh->nud_state & NUD_VALID) &&
411 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
412 struct in6_addr mcaddr;
413 struct in6_addr *target;
415 neigh->updated = jiffies;
416 read_unlock_bh(&neigh->lock);
418 target = (struct in6_addr *)&neigh->primary_key;
419 addrconf_addr_solict_mult(target, &mcaddr);
420 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
422 read_unlock_bh(&neigh->lock);
428 static inline void rt6_probe(struct rt6_info *rt)
434 * Default Router Selection (RFC 2461 6.3.6)
436 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
438 struct net_device *dev = rt->dst.dev;
439 if (!oif || dev->ifindex == oif)
441 if ((dev->flags & IFF_LOOPBACK) &&
442 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
447 static inline int rt6_check_neigh(struct rt6_info *rt)
449 struct neighbour *neigh;
453 neigh = dst_get_neighbour_noref(&rt->dst);
454 if (rt->rt6i_flags & RTF_NONEXTHOP ||
455 !(rt->rt6i_flags & RTF_GATEWAY))
458 read_lock_bh(&neigh->lock);
459 if (neigh->nud_state & NUD_VALID)
461 #ifdef CONFIG_IPV6_ROUTER_PREF
462 else if (neigh->nud_state & NUD_FAILED)
467 read_unlock_bh(&neigh->lock);
474 static int rt6_score_route(struct rt6_info *rt, int oif,
479 m = rt6_check_dev(rt, oif);
480 if (!m && (strict & RT6_LOOKUP_F_IFACE))
482 #ifdef CONFIG_IPV6_ROUTER_PREF
483 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
485 n = rt6_check_neigh(rt);
486 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
491 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
492 int *mpri, struct rt6_info *match)
496 if (rt6_check_expired(rt))
499 m = rt6_score_route(rt, oif, strict);
504 if (strict & RT6_LOOKUP_F_REACHABLE)
508 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
516 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
517 struct rt6_info *rr_head,
518 u32 metric, int oif, int strict)
520 struct rt6_info *rt, *match;
524 for (rt = rr_head; rt && rt->rt6i_metric == metric;
525 rt = rt->dst.rt6_next)
526 match = find_match(rt, oif, strict, &mpri, match);
527 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
528 rt = rt->dst.rt6_next)
529 match = find_match(rt, oif, strict, &mpri, match);
534 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
536 struct rt6_info *match, *rt0;
541 fn->rr_ptr = rt0 = fn->leaf;
543 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
546 (strict & RT6_LOOKUP_F_REACHABLE)) {
547 struct rt6_info *next = rt0->dst.rt6_next;
549 /* no entries matched; do round-robin */
550 if (!next || next->rt6i_metric != rt0->rt6i_metric)
557 net = dev_net(rt0->dst.dev);
558 return match ? match : net->ipv6.ip6_null_entry;
561 #ifdef CONFIG_IPV6_ROUTE_INFO
562 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
563 const struct in6_addr *gwaddr)
565 struct net *net = dev_net(dev);
566 struct route_info *rinfo = (struct route_info *) opt;
567 struct in6_addr prefix_buf, *prefix;
569 unsigned long lifetime;
572 if (len < sizeof(struct route_info)) {
576 /* Sanity check for prefix_len and length */
577 if (rinfo->length > 3) {
579 } else if (rinfo->prefix_len > 128) {
581 } else if (rinfo->prefix_len > 64) {
582 if (rinfo->length < 2) {
585 } else if (rinfo->prefix_len > 0) {
586 if (rinfo->length < 1) {
591 pref = rinfo->route_pref;
592 if (pref == ICMPV6_ROUTER_PREF_INVALID)
595 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
597 if (rinfo->length == 3)
598 prefix = (struct in6_addr *)rinfo->prefix;
600 /* this function is safe */
601 ipv6_addr_prefix(&prefix_buf,
602 (struct in6_addr *)rinfo->prefix,
604 prefix = &prefix_buf;
607 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
610 if (rt && !lifetime) {
616 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
619 rt->rt6i_flags = RTF_ROUTEINFO |
620 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
623 if (!addrconf_finite_timeout(lifetime)) {
624 rt->rt6i_flags &= ~RTF_EXPIRES;
626 rt->dst.expires = jiffies + HZ * lifetime;
627 rt->rt6i_flags |= RTF_EXPIRES;
629 dst_release(&rt->dst);
635 #define BACKTRACK(__net, saddr) \
637 if (rt == __net->ipv6.ip6_null_entry) { \
638 struct fib6_node *pn; \
640 if (fn->fn_flags & RTN_TL_ROOT) \
643 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
644 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
647 if (fn->fn_flags & RTN_RTINFO) \
653 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
654 struct fib6_table *table,
655 struct flowi6 *fl6, int flags)
657 struct fib6_node *fn;
660 read_lock_bh(&table->tb6_lock);
661 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
664 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
665 BACKTRACK(net, &fl6->saddr);
667 dst_use(&rt->dst, jiffies);
668 read_unlock_bh(&table->tb6_lock);
673 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
676 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
678 EXPORT_SYMBOL_GPL(ip6_route_lookup);
680 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
681 const struct in6_addr *saddr, int oif, int strict)
683 struct flowi6 fl6 = {
687 struct dst_entry *dst;
688 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
691 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
692 flags |= RT6_LOOKUP_F_HAS_SADDR;
695 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
697 return (struct rt6_info *) dst;
704 EXPORT_SYMBOL(rt6_lookup);
706 /* ip6_ins_rt is called with FREE table->tb6_lock.
707 It takes new route entry, the addition fails by any reason the
708 route is freed. In any case, if caller does not hold it, it may
712 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
715 struct fib6_table *table;
717 table = rt->rt6i_table;
718 write_lock_bh(&table->tb6_lock);
719 err = fib6_add(&table->tb6_root, rt, info);
720 write_unlock_bh(&table->tb6_lock);
725 int ip6_ins_rt(struct rt6_info *rt)
727 struct nl_info info = {
728 .nl_net = dev_net(rt->dst.dev),
730 return __ip6_ins_rt(rt, &info);
733 static struct rt6_info *rt6_alloc_cow(const struct rt6_info *ort,
734 const struct in6_addr *daddr,
735 const struct in6_addr *saddr)
743 rt = ip6_rt_copy(ort, daddr);
746 int attempts = !in_softirq();
748 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
749 if (ort->rt6i_dst.plen != 128 &&
750 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
751 rt->rt6i_flags |= RTF_ANYCAST;
752 rt->rt6i_gateway = *daddr;
755 rt->rt6i_flags |= RTF_CACHE;
757 #ifdef CONFIG_IPV6_SUBTREES
758 if (rt->rt6i_src.plen && saddr) {
759 rt->rt6i_src.addr = *saddr;
760 rt->rt6i_src.plen = 128;
765 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
766 struct net *net = dev_net(rt->dst.dev);
767 int saved_rt_min_interval =
768 net->ipv6.sysctl.ip6_rt_gc_min_interval;
769 int saved_rt_elasticity =
770 net->ipv6.sysctl.ip6_rt_gc_elasticity;
772 if (attempts-- > 0) {
773 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
774 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
776 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
778 net->ipv6.sysctl.ip6_rt_gc_elasticity =
780 net->ipv6.sysctl.ip6_rt_gc_min_interval =
781 saved_rt_min_interval;
787 "ipv6: Neighbour table overflow.\n");
796 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
797 const struct in6_addr *daddr)
799 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
802 rt->rt6i_flags |= RTF_CACHE;
803 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_noref_raw(&ort->dst)));
808 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
809 struct flowi6 *fl6, int flags)
811 struct fib6_node *fn;
812 struct rt6_info *rt, *nrt;
816 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
818 strict |= flags & RT6_LOOKUP_F_IFACE;
821 read_lock_bh(&table->tb6_lock);
824 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
827 rt = rt6_select(fn, oif, strict | reachable);
829 BACKTRACK(net, &fl6->saddr);
830 if (rt == net->ipv6.ip6_null_entry ||
831 rt->rt6i_flags & RTF_CACHE)
835 read_unlock_bh(&table->tb6_lock);
837 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
838 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
839 else if (!(rt->dst.flags & DST_HOST))
840 nrt = rt6_alloc_clone(rt, &fl6->daddr);
844 dst_release(&rt->dst);
845 rt = nrt ? : net->ipv6.ip6_null_entry;
849 err = ip6_ins_rt(nrt);
858 * Race condition! In the gap, when table->tb6_lock was
859 * released someone could insert this route. Relookup.
861 dst_release(&rt->dst);
870 read_unlock_bh(&table->tb6_lock);
872 rt->dst.lastuse = jiffies;
878 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
879 struct flowi6 *fl6, int flags)
881 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
884 void ip6_route_input(struct sk_buff *skb)
886 const struct ipv6hdr *iph = ipv6_hdr(skb);
887 struct net *net = dev_net(skb->dev);
888 int flags = RT6_LOOKUP_F_HAS_SADDR;
889 struct flowi6 fl6 = {
890 .flowi6_iif = skb->dev->ifindex,
893 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
894 .flowi6_mark = skb->mark,
895 .flowi6_proto = iph->nexthdr,
898 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
899 flags |= RT6_LOOKUP_F_IFACE;
901 skb_dst_set(skb, fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_input));
904 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
905 struct flowi6 *fl6, int flags)
907 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
910 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
915 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
916 flags |= RT6_LOOKUP_F_IFACE;
918 if (!ipv6_addr_any(&fl6->saddr))
919 flags |= RT6_LOOKUP_F_HAS_SADDR;
921 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
923 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
926 EXPORT_SYMBOL(ip6_route_output);
928 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
930 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
931 struct dst_entry *new = NULL;
933 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
935 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
940 new->input = dst_discard;
941 new->output = dst_discard;
943 if (dst_metrics_read_only(&ort->dst))
944 new->_metrics = ort->dst._metrics;
946 dst_copy_metrics(new, &ort->dst);
947 rt->rt6i_idev = ort->rt6i_idev;
949 in6_dev_hold(rt->rt6i_idev);
952 rt->rt6i_gateway = ort->rt6i_gateway;
953 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
956 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
957 #ifdef CONFIG_IPV6_SUBTREES
958 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
964 dst_release(dst_orig);
965 return new ? new : ERR_PTR(-ENOMEM);
969 * Destination cache support functions
972 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
976 rt = (struct rt6_info *) dst;
978 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
979 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
981 rt6_bind_peer(rt, 0);
982 rt->rt6i_peer_genid = rt6_peer_genid();
989 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
991 struct rt6_info *rt = (struct rt6_info *) dst;
994 if (rt->rt6i_flags & RTF_CACHE) {
995 if (rt6_check_expired(rt)) {
1007 static void ip6_link_failure(struct sk_buff *skb)
1009 struct rt6_info *rt;
1011 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1013 rt = (struct rt6_info *) skb_dst(skb);
1015 if (rt->rt6i_flags & RTF_CACHE) {
1016 dst_set_expires(&rt->dst, 0);
1017 rt->rt6i_flags |= RTF_EXPIRES;
1018 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1019 rt->rt6i_node->fn_sernum = -1;
1023 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1025 struct rt6_info *rt6 = (struct rt6_info*)dst;
1027 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1028 rt6->rt6i_flags |= RTF_MODIFIED;
1029 if (mtu < IPV6_MIN_MTU) {
1030 u32 features = dst_metric(dst, RTAX_FEATURES);
1032 features |= RTAX_FEATURE_ALLFRAG;
1033 dst_metric_set(dst, RTAX_FEATURES, features);
1035 dst_metric_set(dst, RTAX_MTU, mtu);
1039 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1041 struct net_device *dev = dst->dev;
1042 unsigned int mtu = dst_mtu(dst);
1043 struct net *net = dev_net(dev);
1045 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1047 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1048 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1051 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1052 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1053 * IPV6_MAXPLEN is also valid and means: "any MSS,
1054 * rely only on pmtu discovery"
1056 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1061 static unsigned int ip6_mtu(const struct dst_entry *dst)
1063 struct inet6_dev *idev;
1064 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1072 idev = __in6_dev_get(dst->dev);
1074 mtu = idev->cnf.mtu6;
1080 static struct dst_entry *icmp6_dst_gc_list;
1081 static DEFINE_SPINLOCK(icmp6_dst_lock);
1083 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1084 struct neighbour *neigh,
1087 struct dst_entry *dst;
1088 struct rt6_info *rt;
1089 struct inet6_dev *idev = in6_dev_get(dev);
1090 struct net *net = dev_net(dev);
1092 if (unlikely(!idev))
1093 return ERR_PTR(-ENODEV);
1095 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, dev, 0);
1096 if (unlikely(!rt)) {
1098 dst = ERR_PTR(-ENOMEM);
1105 neigh = ip6_neigh_lookup(&rt->dst, &fl6->daddr);
1106 if (IS_ERR(neigh)) {
1109 return ERR_CAST(neigh);
1113 rt->dst.flags |= DST_HOST;
1114 rt->dst.output = ip6_output;
1115 dst_set_neighbour(&rt->dst, neigh);
1116 atomic_set(&rt->dst.__refcnt, 1);
1117 rt->rt6i_dst.addr = fl6->daddr;
1118 rt->rt6i_dst.plen = 128;
1119 rt->rt6i_idev = idev;
1120 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1122 spin_lock_bh(&icmp6_dst_lock);
1123 rt->dst.next = icmp6_dst_gc_list;
1124 icmp6_dst_gc_list = &rt->dst;
1125 spin_unlock_bh(&icmp6_dst_lock);
1127 fib6_force_start_gc(net);
1129 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1135 int icmp6_dst_gc(void)
1137 struct dst_entry *dst, **pprev;
1140 spin_lock_bh(&icmp6_dst_lock);
1141 pprev = &icmp6_dst_gc_list;
1143 while ((dst = *pprev) != NULL) {
1144 if (!atomic_read(&dst->__refcnt)) {
1153 spin_unlock_bh(&icmp6_dst_lock);
1158 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1161 struct dst_entry *dst, **pprev;
1163 spin_lock_bh(&icmp6_dst_lock);
1164 pprev = &icmp6_dst_gc_list;
1165 while ((dst = *pprev) != NULL) {
1166 struct rt6_info *rt = (struct rt6_info *) dst;
1167 if (func(rt, arg)) {
1174 spin_unlock_bh(&icmp6_dst_lock);
1177 static int ip6_dst_gc(struct dst_ops *ops)
1179 unsigned long now = jiffies;
1180 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1181 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1182 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1183 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1184 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1185 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1188 entries = dst_entries_get_fast(ops);
1189 if (time_after(rt_last_gc + rt_min_interval, now) &&
1190 entries <= rt_max_size)
1193 net->ipv6.ip6_rt_gc_expire++;
1194 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1195 net->ipv6.ip6_rt_last_gc = now;
1196 entries = dst_entries_get_slow(ops);
1197 if (entries < ops->gc_thresh)
1198 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1200 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1201 return entries > rt_max_size;
1204 /* Clean host part of a prefix. Not necessary in radix tree,
1205 but results in cleaner routing tables.
1207 Remove it only when all the things will work!
1210 int ip6_dst_hoplimit(struct dst_entry *dst)
1212 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1213 if (hoplimit == 0) {
1214 struct net_device *dev = dst->dev;
1215 struct inet6_dev *idev;
1218 idev = __in6_dev_get(dev);
1220 hoplimit = idev->cnf.hop_limit;
1222 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1227 EXPORT_SYMBOL(ip6_dst_hoplimit);
1233 int ip6_route_add(struct fib6_config *cfg)
1236 struct net *net = cfg->fc_nlinfo.nl_net;
1237 struct rt6_info *rt = NULL;
1238 struct net_device *dev = NULL;
1239 struct inet6_dev *idev = NULL;
1240 struct fib6_table *table;
1243 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1245 #ifndef CONFIG_IPV6_SUBTREES
1246 if (cfg->fc_src_len)
1249 if (cfg->fc_ifindex) {
1251 dev = dev_get_by_index(net, cfg->fc_ifindex);
1254 idev = in6_dev_get(dev);
1259 if (cfg->fc_metric == 0)
1260 cfg->fc_metric = IP6_RT_PRIO_USER;
1263 if (cfg->fc_nlinfo.nlh &&
1264 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1265 table = fib6_get_table(net, cfg->fc_table);
1267 printk(KERN_WARNING "IPv6: NLM_F_CREATE should be specified when creating new route\n");
1268 table = fib6_new_table(net, cfg->fc_table);
1271 table = fib6_new_table(net, cfg->fc_table);
1277 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops, NULL, DST_NOCOUNT);
1284 rt->dst.obsolete = -1;
1285 rt->dst.expires = (cfg->fc_flags & RTF_EXPIRES) ?
1286 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1289 if (cfg->fc_protocol == RTPROT_UNSPEC)
1290 cfg->fc_protocol = RTPROT_BOOT;
1291 rt->rt6i_protocol = cfg->fc_protocol;
1293 addr_type = ipv6_addr_type(&cfg->fc_dst);
1295 if (addr_type & IPV6_ADDR_MULTICAST)
1296 rt->dst.input = ip6_mc_input;
1297 else if (cfg->fc_flags & RTF_LOCAL)
1298 rt->dst.input = ip6_input;
1300 rt->dst.input = ip6_forward;
1302 rt->dst.output = ip6_output;
1304 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1305 rt->rt6i_dst.plen = cfg->fc_dst_len;
1306 if (rt->rt6i_dst.plen == 128)
1307 rt->dst.flags |= DST_HOST;
1309 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1310 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1315 dst_init_metrics(&rt->dst, metrics, 0);
1317 #ifdef CONFIG_IPV6_SUBTREES
1318 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1319 rt->rt6i_src.plen = cfg->fc_src_len;
1322 rt->rt6i_metric = cfg->fc_metric;
1324 /* We cannot add true routes via loopback here,
1325 they would result in kernel looping; promote them to reject routes
1327 if ((cfg->fc_flags & RTF_REJECT) ||
1328 (dev && (dev->flags & IFF_LOOPBACK) &&
1329 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1330 !(cfg->fc_flags & RTF_LOCAL))) {
1331 /* hold loopback dev/idev if we haven't done so. */
1332 if (dev != net->loopback_dev) {
1337 dev = net->loopback_dev;
1339 idev = in6_dev_get(dev);
1345 rt->dst.output = ip6_pkt_discard_out;
1346 rt->dst.input = ip6_pkt_discard;
1347 rt->dst.error = -ENETUNREACH;
1348 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1352 if (cfg->fc_flags & RTF_GATEWAY) {
1353 const struct in6_addr *gw_addr;
1356 gw_addr = &cfg->fc_gateway;
1357 rt->rt6i_gateway = *gw_addr;
1358 gwa_type = ipv6_addr_type(gw_addr);
1360 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1361 struct rt6_info *grt;
1363 /* IPv6 strictly inhibits using not link-local
1364 addresses as nexthop address.
1365 Otherwise, router will not able to send redirects.
1366 It is very good, but in some (rare!) circumstances
1367 (SIT, PtP, NBMA NOARP links) it is handy to allow
1368 some exceptions. --ANK
1371 if (!(gwa_type & IPV6_ADDR_UNICAST))
1374 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1376 err = -EHOSTUNREACH;
1380 if (dev != grt->dst.dev) {
1381 dst_release(&grt->dst);
1386 idev = grt->rt6i_idev;
1388 in6_dev_hold(grt->rt6i_idev);
1390 if (!(grt->rt6i_flags & RTF_GATEWAY))
1392 dst_release(&grt->dst);
1398 if (!dev || (dev->flags & IFF_LOOPBACK))
1406 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1407 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1411 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1412 rt->rt6i_prefsrc.plen = 128;
1414 rt->rt6i_prefsrc.plen = 0;
1416 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1417 err = rt6_bind_neighbour(rt, dev);
1422 rt->rt6i_flags = cfg->fc_flags;
1429 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1430 int type = nla_type(nla);
1433 if (type > RTAX_MAX) {
1438 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1444 rt->rt6i_idev = idev;
1445 rt->rt6i_table = table;
1447 cfg->fc_nlinfo.nl_net = dev_net(dev);
1449 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1461 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1464 struct fib6_table *table;
1465 struct net *net = dev_net(rt->dst.dev);
1467 if (rt == net->ipv6.ip6_null_entry)
1470 table = rt->rt6i_table;
1471 write_lock_bh(&table->tb6_lock);
1473 err = fib6_del(rt, info);
1474 dst_release(&rt->dst);
1476 write_unlock_bh(&table->tb6_lock);
1481 int ip6_del_rt(struct rt6_info *rt)
1483 struct nl_info info = {
1484 .nl_net = dev_net(rt->dst.dev),
1486 return __ip6_del_rt(rt, &info);
1489 static int ip6_route_del(struct fib6_config *cfg)
1491 struct fib6_table *table;
1492 struct fib6_node *fn;
1493 struct rt6_info *rt;
1496 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1500 read_lock_bh(&table->tb6_lock);
1502 fn = fib6_locate(&table->tb6_root,
1503 &cfg->fc_dst, cfg->fc_dst_len,
1504 &cfg->fc_src, cfg->fc_src_len);
1507 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1508 if (cfg->fc_ifindex &&
1510 rt->dst.dev->ifindex != cfg->fc_ifindex))
1512 if (cfg->fc_flags & RTF_GATEWAY &&
1513 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1515 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1518 read_unlock_bh(&table->tb6_lock);
1520 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1523 read_unlock_bh(&table->tb6_lock);
1531 struct ip6rd_flowi {
1533 struct in6_addr gateway;
1536 static struct rt6_info *__ip6_route_redirect(struct net *net,
1537 struct fib6_table *table,
1541 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1542 struct rt6_info *rt;
1543 struct fib6_node *fn;
1546 * Get the "current" route for this destination and
1547 * check if the redirect has come from approriate router.
1549 * RFC 2461 specifies that redirects should only be
1550 * accepted if they come from the nexthop to the target.
1551 * Due to the way the routes are chosen, this notion
1552 * is a bit fuzzy and one might need to check all possible
1556 read_lock_bh(&table->tb6_lock);
1557 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1559 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1561 * Current route is on-link; redirect is always invalid.
1563 * Seems, previous statement is not true. It could
1564 * be node, which looks for us as on-link (f.e. proxy ndisc)
1565 * But then router serving it might decide, that we should
1566 * know truth 8)8) --ANK (980726).
1568 if (rt6_check_expired(rt))
1570 if (!(rt->rt6i_flags & RTF_GATEWAY))
1572 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1574 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1580 rt = net->ipv6.ip6_null_entry;
1581 BACKTRACK(net, &fl6->saddr);
1585 read_unlock_bh(&table->tb6_lock);
1590 static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1591 const struct in6_addr *src,
1592 const struct in6_addr *gateway,
1593 struct net_device *dev)
1595 int flags = RT6_LOOKUP_F_HAS_SADDR;
1596 struct net *net = dev_net(dev);
1597 struct ip6rd_flowi rdfl = {
1599 .flowi6_oif = dev->ifindex,
1605 rdfl.gateway = *gateway;
1607 if (rt6_need_strict(dest))
1608 flags |= RT6_LOOKUP_F_IFACE;
1610 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1611 flags, __ip6_route_redirect);
1614 void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1615 const struct in6_addr *saddr,
1616 struct neighbour *neigh, u8 *lladdr, int on_link)
1618 struct rt6_info *rt, *nrt = NULL;
1619 struct netevent_redirect netevent;
1620 struct net *net = dev_net(neigh->dev);
1622 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1624 if (rt == net->ipv6.ip6_null_entry) {
1625 if (net_ratelimit())
1626 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1627 "for redirect target\n");
1632 * We have finally decided to accept it.
1635 neigh_update(neigh, lladdr, NUD_STALE,
1636 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1637 NEIGH_UPDATE_F_OVERRIDE|
1638 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1639 NEIGH_UPDATE_F_ISROUTER))
1643 * Redirect received -> path was valid.
1644 * Look, redirects are sent only in response to data packets,
1645 * so that this nexthop apparently is reachable. --ANK
1647 dst_confirm(&rt->dst);
1649 /* Duplicate redirect: silently ignore. */
1650 if (neigh == dst_get_neighbour_noref_raw(&rt->dst))
1653 nrt = ip6_rt_copy(rt, dest);
1657 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1659 nrt->rt6i_flags &= ~RTF_GATEWAY;
1661 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1662 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1664 if (ip6_ins_rt(nrt))
1667 netevent.old = &rt->dst;
1668 netevent.new = &nrt->dst;
1669 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1671 if (rt->rt6i_flags & RTF_CACHE) {
1677 dst_release(&rt->dst);
1681 * Handle ICMP "packet too big" messages
1682 * i.e. Path MTU discovery
1685 static void rt6_do_pmtu_disc(const struct in6_addr *daddr, const struct in6_addr *saddr,
1686 struct net *net, u32 pmtu, int ifindex)
1688 struct rt6_info *rt, *nrt;
1691 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1695 if (rt6_check_expired(rt)) {
1700 if (pmtu >= dst_mtu(&rt->dst))
1703 if (pmtu < IPV6_MIN_MTU) {
1705 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1706 * MTU (1280) and a fragment header should always be included
1707 * after a node receiving Too Big message reporting PMTU is
1708 * less than the IPv6 Minimum Link MTU.
1710 pmtu = IPV6_MIN_MTU;
1714 /* New mtu received -> path was valid.
1715 They are sent only in response to data packets,
1716 so that this nexthop apparently is reachable. --ANK
1718 dst_confirm(&rt->dst);
1720 /* Host route. If it is static, it would be better
1721 not to override it, but add new one, so that
1722 when cache entry will expire old pmtu
1723 would return automatically.
1725 if (rt->rt6i_flags & RTF_CACHE) {
1726 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1728 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1729 features |= RTAX_FEATURE_ALLFRAG;
1730 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1732 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1733 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1738 Two cases are possible:
1739 1. It is connected route. Action: COW
1740 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1742 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
1743 nrt = rt6_alloc_cow(rt, daddr, saddr);
1745 nrt = rt6_alloc_clone(rt, daddr);
1748 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1750 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1751 features |= RTAX_FEATURE_ALLFRAG;
1752 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1755 /* According to RFC 1981, detecting PMTU increase shouldn't be
1756 * happened within 5 mins, the recommended timer is 10 mins.
1757 * Here this route expiration time is set to ip6_rt_mtu_expires
1758 * which is 10 mins. After 10 mins the decreased pmtu is expired
1759 * and detecting PMTU increase will be automatically happened.
1761 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1762 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1767 dst_release(&rt->dst);
1770 void rt6_pmtu_discovery(const struct in6_addr *daddr, const struct in6_addr *saddr,
1771 struct net_device *dev, u32 pmtu)
1773 struct net *net = dev_net(dev);
1776 * RFC 1981 states that a node "MUST reduce the size of the packets it
1777 * is sending along the path" that caused the Packet Too Big message.
1778 * Since it's not possible in the general case to determine which
1779 * interface was used to send the original packet, we update the MTU
1780 * on the interface that will be used to send future packets. We also
1781 * update the MTU on the interface that received the Packet Too Big in
1782 * case the original packet was forced out that interface with
1783 * SO_BINDTODEVICE or similar. This is the next best thing to the
1784 * correct behaviour, which would be to update the MTU on all
1787 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1788 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1792 * Misc support functions
1795 static struct rt6_info *ip6_rt_copy(const struct rt6_info *ort,
1796 const struct in6_addr *dest)
1798 struct net *net = dev_net(ort->dst.dev);
1799 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
1803 rt->dst.input = ort->dst.input;
1804 rt->dst.output = ort->dst.output;
1805 rt->dst.flags |= DST_HOST;
1807 rt->rt6i_dst.addr = *dest;
1808 rt->rt6i_dst.plen = 128;
1809 dst_copy_metrics(&rt->dst, &ort->dst);
1810 rt->dst.error = ort->dst.error;
1811 rt->rt6i_idev = ort->rt6i_idev;
1813 in6_dev_hold(rt->rt6i_idev);
1814 rt->dst.lastuse = jiffies;
1815 rt->dst.expires = 0;
1817 rt->rt6i_gateway = ort->rt6i_gateway;
1818 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1819 rt->rt6i_metric = 0;
1821 #ifdef CONFIG_IPV6_SUBTREES
1822 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1824 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1825 rt->rt6i_table = ort->rt6i_table;
1830 #ifdef CONFIG_IPV6_ROUTE_INFO
1831 static struct rt6_info *rt6_get_route_info(struct net *net,
1832 const struct in6_addr *prefix, int prefixlen,
1833 const struct in6_addr *gwaddr, int ifindex)
1835 struct fib6_node *fn;
1836 struct rt6_info *rt = NULL;
1837 struct fib6_table *table;
1839 table = fib6_get_table(net, RT6_TABLE_INFO);
1843 write_lock_bh(&table->tb6_lock);
1844 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1848 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1849 if (rt->dst.dev->ifindex != ifindex)
1851 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1853 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1859 write_unlock_bh(&table->tb6_lock);
1863 static struct rt6_info *rt6_add_route_info(struct net *net,
1864 const struct in6_addr *prefix, int prefixlen,
1865 const struct in6_addr *gwaddr, int ifindex,
1868 struct fib6_config cfg = {
1869 .fc_table = RT6_TABLE_INFO,
1870 .fc_metric = IP6_RT_PRIO_USER,
1871 .fc_ifindex = ifindex,
1872 .fc_dst_len = prefixlen,
1873 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1874 RTF_UP | RTF_PREF(pref),
1876 .fc_nlinfo.nlh = NULL,
1877 .fc_nlinfo.nl_net = net,
1880 cfg.fc_dst = *prefix;
1881 cfg.fc_gateway = *gwaddr;
1883 /* We should treat it as a default route if prefix length is 0. */
1885 cfg.fc_flags |= RTF_DEFAULT;
1887 ip6_route_add(&cfg);
1889 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1893 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1895 struct rt6_info *rt;
1896 struct fib6_table *table;
1898 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1902 write_lock_bh(&table->tb6_lock);
1903 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1904 if (dev == rt->dst.dev &&
1905 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1906 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1911 write_unlock_bh(&table->tb6_lock);
1915 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1916 struct net_device *dev,
1919 struct fib6_config cfg = {
1920 .fc_table = RT6_TABLE_DFLT,
1921 .fc_metric = IP6_RT_PRIO_USER,
1922 .fc_ifindex = dev->ifindex,
1923 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1924 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1926 .fc_nlinfo.nlh = NULL,
1927 .fc_nlinfo.nl_net = dev_net(dev),
1930 cfg.fc_gateway = *gwaddr;
1932 ip6_route_add(&cfg);
1934 return rt6_get_dflt_router(gwaddr, dev);
1937 void rt6_purge_dflt_routers(struct net *net)
1939 struct rt6_info *rt;
1940 struct fib6_table *table;
1942 /* NOTE: Keep consistent with rt6_get_dflt_router */
1943 table = fib6_get_table(net, RT6_TABLE_DFLT);
1948 read_lock_bh(&table->tb6_lock);
1949 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1950 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1952 read_unlock_bh(&table->tb6_lock);
1957 read_unlock_bh(&table->tb6_lock);
1960 static void rtmsg_to_fib6_config(struct net *net,
1961 struct in6_rtmsg *rtmsg,
1962 struct fib6_config *cfg)
1964 memset(cfg, 0, sizeof(*cfg));
1966 cfg->fc_table = RT6_TABLE_MAIN;
1967 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1968 cfg->fc_metric = rtmsg->rtmsg_metric;
1969 cfg->fc_expires = rtmsg->rtmsg_info;
1970 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1971 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1972 cfg->fc_flags = rtmsg->rtmsg_flags;
1974 cfg->fc_nlinfo.nl_net = net;
1976 cfg->fc_dst = rtmsg->rtmsg_dst;
1977 cfg->fc_src = rtmsg->rtmsg_src;
1978 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1981 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1983 struct fib6_config cfg;
1984 struct in6_rtmsg rtmsg;
1988 case SIOCADDRT: /* Add a route */
1989 case SIOCDELRT: /* Delete a route */
1990 if (!capable(CAP_NET_ADMIN))
1992 err = copy_from_user(&rtmsg, arg,
1993 sizeof(struct in6_rtmsg));
1997 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2002 err = ip6_route_add(&cfg);
2005 err = ip6_route_del(&cfg);
2019 * Drop the packet on the floor
2022 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2025 struct dst_entry *dst = skb_dst(skb);
2026 switch (ipstats_mib_noroutes) {
2027 case IPSTATS_MIB_INNOROUTES:
2028 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2029 if (type == IPV6_ADDR_ANY) {
2030 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2031 IPSTATS_MIB_INADDRERRORS);
2035 case IPSTATS_MIB_OUTNOROUTES:
2036 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2037 ipstats_mib_noroutes);
2040 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2045 static int ip6_pkt_discard(struct sk_buff *skb)
2047 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2050 static int ip6_pkt_discard_out(struct sk_buff *skb)
2052 skb->dev = skb_dst(skb)->dev;
2053 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2056 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2058 static int ip6_pkt_prohibit(struct sk_buff *skb)
2060 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2063 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2065 skb->dev = skb_dst(skb)->dev;
2066 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2072 * Allocate a dst for local (unicast / anycast) address.
2075 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2076 const struct in6_addr *addr,
2079 struct net *net = dev_net(idev->dev);
2080 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops,
2081 net->loopback_dev, 0);
2085 if (net_ratelimit())
2086 pr_warning("IPv6: Maximum number of routes reached,"
2087 " consider increasing route/max_size.\n");
2088 return ERR_PTR(-ENOMEM);
2093 rt->dst.flags |= DST_HOST;
2094 rt->dst.input = ip6_input;
2095 rt->dst.output = ip6_output;
2096 rt->rt6i_idev = idev;
2097 rt->dst.obsolete = -1;
2099 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2101 rt->rt6i_flags |= RTF_ANYCAST;
2103 rt->rt6i_flags |= RTF_LOCAL;
2104 err = rt6_bind_neighbour(rt, rt->dst.dev);
2107 return ERR_PTR(err);
2110 rt->rt6i_dst.addr = *addr;
2111 rt->rt6i_dst.plen = 128;
2112 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2114 atomic_set(&rt->dst.__refcnt, 1);
2119 int ip6_route_get_saddr(struct net *net,
2120 struct rt6_info *rt,
2121 const struct in6_addr *daddr,
2123 struct in6_addr *saddr)
2125 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2127 if (rt->rt6i_prefsrc.plen)
2128 *saddr = rt->rt6i_prefsrc.addr;
2130 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2131 daddr, prefs, saddr);
2135 /* remove deleted ip from prefsrc entries */
2136 struct arg_dev_net_ip {
2137 struct net_device *dev;
2139 struct in6_addr *addr;
2142 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2144 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2145 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2146 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2148 if (((void *)rt->dst.dev == dev || !dev) &&
2149 rt != net->ipv6.ip6_null_entry &&
2150 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2151 /* remove prefsrc entry */
2152 rt->rt6i_prefsrc.plen = 0;
2157 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2159 struct net *net = dev_net(ifp->idev->dev);
2160 struct arg_dev_net_ip adni = {
2161 .dev = ifp->idev->dev,
2165 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2168 struct arg_dev_net {
2169 struct net_device *dev;
2173 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2175 const struct arg_dev_net *adn = arg;
2176 const struct net_device *dev = adn->dev;
2178 if ((rt->dst.dev == dev || !dev) &&
2179 rt != adn->net->ipv6.ip6_null_entry)
2185 void rt6_ifdown(struct net *net, struct net_device *dev)
2187 struct arg_dev_net adn = {
2192 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2193 icmp6_clean_all(fib6_ifdown, &adn);
2196 struct rt6_mtu_change_arg
2198 struct net_device *dev;
2202 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2204 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2205 struct inet6_dev *idev;
2207 /* In IPv6 pmtu discovery is not optional,
2208 so that RTAX_MTU lock cannot disable it.
2209 We still use this lock to block changes
2210 caused by addrconf/ndisc.
2213 idev = __in6_dev_get(arg->dev);
2217 /* For administrative MTU increase, there is no way to discover
2218 IPv6 PMTU increase, so PMTU increase should be updated here.
2219 Since RFC 1981 doesn't include administrative MTU increase
2220 update PMTU increase is a MUST. (i.e. jumbo frame)
2223 If new MTU is less than route PMTU, this new MTU will be the
2224 lowest MTU in the path, update the route PMTU to reflect PMTU
2225 decreases; if new MTU is greater than route PMTU, and the
2226 old MTU is the lowest MTU in the path, update the route PMTU
2227 to reflect the increase. In this case if the other nodes' MTU
2228 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2231 if (rt->dst.dev == arg->dev &&
2232 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2233 (dst_mtu(&rt->dst) >= arg->mtu ||
2234 (dst_mtu(&rt->dst) < arg->mtu &&
2235 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2236 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2241 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2243 struct rt6_mtu_change_arg arg = {
2248 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2251 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2252 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2253 [RTA_OIF] = { .type = NLA_U32 },
2254 [RTA_IIF] = { .type = NLA_U32 },
2255 [RTA_PRIORITY] = { .type = NLA_U32 },
2256 [RTA_METRICS] = { .type = NLA_NESTED },
2259 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2260 struct fib6_config *cfg)
2263 struct nlattr *tb[RTA_MAX+1];
2266 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2271 rtm = nlmsg_data(nlh);
2272 memset(cfg, 0, sizeof(*cfg));
2274 cfg->fc_table = rtm->rtm_table;
2275 cfg->fc_dst_len = rtm->rtm_dst_len;
2276 cfg->fc_src_len = rtm->rtm_src_len;
2277 cfg->fc_flags = RTF_UP;
2278 cfg->fc_protocol = rtm->rtm_protocol;
2280 if (rtm->rtm_type == RTN_UNREACHABLE)
2281 cfg->fc_flags |= RTF_REJECT;
2283 if (rtm->rtm_type == RTN_LOCAL)
2284 cfg->fc_flags |= RTF_LOCAL;
2286 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2287 cfg->fc_nlinfo.nlh = nlh;
2288 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2290 if (tb[RTA_GATEWAY]) {
2291 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2292 cfg->fc_flags |= RTF_GATEWAY;
2296 int plen = (rtm->rtm_dst_len + 7) >> 3;
2298 if (nla_len(tb[RTA_DST]) < plen)
2301 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2305 int plen = (rtm->rtm_src_len + 7) >> 3;
2307 if (nla_len(tb[RTA_SRC]) < plen)
2310 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2313 if (tb[RTA_PREFSRC])
2314 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2317 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2319 if (tb[RTA_PRIORITY])
2320 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2322 if (tb[RTA_METRICS]) {
2323 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2324 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2328 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2335 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2337 struct fib6_config cfg;
2340 err = rtm_to_fib6_config(skb, nlh, &cfg);
2344 return ip6_route_del(&cfg);
2347 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2349 struct fib6_config cfg;
2352 err = rtm_to_fib6_config(skb, nlh, &cfg);
2356 return ip6_route_add(&cfg);
2359 static inline size_t rt6_nlmsg_size(void)
2361 return NLMSG_ALIGN(sizeof(struct rtmsg))
2362 + nla_total_size(16) /* RTA_SRC */
2363 + nla_total_size(16) /* RTA_DST */
2364 + nla_total_size(16) /* RTA_GATEWAY */
2365 + nla_total_size(16) /* RTA_PREFSRC */
2366 + nla_total_size(4) /* RTA_TABLE */
2367 + nla_total_size(4) /* RTA_IIF */
2368 + nla_total_size(4) /* RTA_OIF */
2369 + nla_total_size(4) /* RTA_PRIORITY */
2370 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2371 + nla_total_size(sizeof(struct rta_cacheinfo));
2374 static int rt6_fill_node(struct net *net,
2375 struct sk_buff *skb, struct rt6_info *rt,
2376 struct in6_addr *dst, struct in6_addr *src,
2377 int iif, int type, u32 pid, u32 seq,
2378 int prefix, int nowait, unsigned int flags)
2380 const struct inet_peer *peer;
2382 struct nlmsghdr *nlh;
2385 struct neighbour *n;
2388 if (prefix) { /* user wants prefix routes only */
2389 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2390 /* success since this is not a prefix route */
2395 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2399 rtm = nlmsg_data(nlh);
2400 rtm->rtm_family = AF_INET6;
2401 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2402 rtm->rtm_src_len = rt->rt6i_src.plen;
2405 table = rt->rt6i_table->tb6_id;
2407 table = RT6_TABLE_UNSPEC;
2408 rtm->rtm_table = table;
2409 NLA_PUT_U32(skb, RTA_TABLE, table);
2410 if (rt->rt6i_flags & RTF_REJECT)
2411 rtm->rtm_type = RTN_UNREACHABLE;
2412 else if (rt->rt6i_flags & RTF_LOCAL)
2413 rtm->rtm_type = RTN_LOCAL;
2414 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2415 rtm->rtm_type = RTN_LOCAL;
2417 rtm->rtm_type = RTN_UNICAST;
2419 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2420 rtm->rtm_protocol = rt->rt6i_protocol;
2421 if (rt->rt6i_flags & RTF_DYNAMIC)
2422 rtm->rtm_protocol = RTPROT_REDIRECT;
2423 else if (rt->rt6i_flags & RTF_ADDRCONF)
2424 rtm->rtm_protocol = RTPROT_KERNEL;
2425 else if (rt->rt6i_flags & RTF_DEFAULT)
2426 rtm->rtm_protocol = RTPROT_RA;
2428 if (rt->rt6i_flags & RTF_CACHE)
2429 rtm->rtm_flags |= RTM_F_CLONED;
2432 NLA_PUT(skb, RTA_DST, 16, dst);
2433 rtm->rtm_dst_len = 128;
2434 } else if (rtm->rtm_dst_len)
2435 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2436 #ifdef CONFIG_IPV6_SUBTREES
2438 NLA_PUT(skb, RTA_SRC, 16, src);
2439 rtm->rtm_src_len = 128;
2440 } else if (rtm->rtm_src_len)
2441 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2444 #ifdef CONFIG_IPV6_MROUTE
2445 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2446 int err = ip6mr_get_route(net, skb, rtm, nowait);
2451 goto nla_put_failure;
2453 if (err == -EMSGSIZE)
2454 goto nla_put_failure;
2459 NLA_PUT_U32(skb, RTA_IIF, iif);
2461 struct in6_addr saddr_buf;
2462 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0)
2463 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2466 if (rt->rt6i_prefsrc.plen) {
2467 struct in6_addr saddr_buf;
2468 saddr_buf = rt->rt6i_prefsrc.addr;
2469 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2472 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2473 goto nla_put_failure;
2476 n = dst_get_neighbour_noref(&rt->dst);
2478 NLA_PUT(skb, RTA_GATEWAY, 16, &n->primary_key);
2482 NLA_PUT_U32(skb, RTA_OIF, rt->dst.dev->ifindex);
2484 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2486 if (!(rt->rt6i_flags & RTF_EXPIRES))
2488 else if (rt->dst.expires - jiffies < INT_MAX)
2489 expires = rt->dst.expires - jiffies;
2493 peer = rt->rt6i_peer;
2495 if (peer && peer->tcp_ts_stamp) {
2497 tsage = get_seconds() - peer->tcp_ts_stamp;
2500 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, ts, tsage,
2501 expires, rt->dst.error) < 0)
2502 goto nla_put_failure;
2504 return nlmsg_end(skb, nlh);
2507 nlmsg_cancel(skb, nlh);
2511 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2513 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2516 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2517 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2518 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2522 return rt6_fill_node(arg->net,
2523 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2524 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2525 prefix, 0, NLM_F_MULTI);
2528 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2530 struct net *net = sock_net(in_skb->sk);
2531 struct nlattr *tb[RTA_MAX+1];
2532 struct rt6_info *rt;
2533 struct sk_buff *skb;
2538 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2543 memset(&fl6, 0, sizeof(fl6));
2546 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2549 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2553 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2556 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2560 iif = nla_get_u32(tb[RTA_IIF]);
2563 fl6.flowi6_oif = nla_get_u32(tb[RTA_OIF]);
2566 struct net_device *dev;
2567 dev = __dev_get_by_index(net, iif);
2574 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2580 /* Reserve room for dummy headers, this skb can pass
2581 through good chunk of routing engine.
2583 skb_reset_mac_header(skb);
2584 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2586 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl6);
2587 skb_dst_set(skb, &rt->dst);
2589 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2590 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2591 nlh->nlmsg_seq, 0, 0, 0);
2597 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2602 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2604 struct sk_buff *skb;
2605 struct net *net = info->nl_net;
2610 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2612 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2616 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2617 event, info->pid, seq, 0, 0, 0);
2619 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2620 WARN_ON(err == -EMSGSIZE);
2624 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2625 info->nlh, gfp_any());
2629 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2632 static int ip6_route_dev_notify(struct notifier_block *this,
2633 unsigned long event, void *data)
2635 struct net_device *dev = (struct net_device *)data;
2636 struct net *net = dev_net(dev);
2638 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2639 net->ipv6.ip6_null_entry->dst.dev = dev;
2640 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2641 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2642 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2643 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2644 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2645 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2656 #ifdef CONFIG_PROC_FS
2667 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2669 struct seq_file *m = p_arg;
2670 struct neighbour *n;
2672 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2674 #ifdef CONFIG_IPV6_SUBTREES
2675 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2677 seq_puts(m, "00000000000000000000000000000000 00 ");
2680 n = dst_get_neighbour_noref(&rt->dst);
2682 seq_printf(m, "%pi6", n->primary_key);
2684 seq_puts(m, "00000000000000000000000000000000");
2687 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2688 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2689 rt->dst.__use, rt->rt6i_flags,
2690 rt->dst.dev ? rt->dst.dev->name : "");
2694 static int ipv6_route_show(struct seq_file *m, void *v)
2696 struct net *net = (struct net *)m->private;
2697 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2701 static int ipv6_route_open(struct inode *inode, struct file *file)
2703 return single_open_net(inode, file, ipv6_route_show);
2706 static const struct file_operations ipv6_route_proc_fops = {
2707 .owner = THIS_MODULE,
2708 .open = ipv6_route_open,
2710 .llseek = seq_lseek,
2711 .release = single_release_net,
2714 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2716 struct net *net = (struct net *)seq->private;
2717 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2718 net->ipv6.rt6_stats->fib_nodes,
2719 net->ipv6.rt6_stats->fib_route_nodes,
2720 net->ipv6.rt6_stats->fib_rt_alloc,
2721 net->ipv6.rt6_stats->fib_rt_entries,
2722 net->ipv6.rt6_stats->fib_rt_cache,
2723 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2724 net->ipv6.rt6_stats->fib_discarded_routes);
2729 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2731 return single_open_net(inode, file, rt6_stats_seq_show);
2734 static const struct file_operations rt6_stats_seq_fops = {
2735 .owner = THIS_MODULE,
2736 .open = rt6_stats_seq_open,
2738 .llseek = seq_lseek,
2739 .release = single_release_net,
2741 #endif /* CONFIG_PROC_FS */
2743 #ifdef CONFIG_SYSCTL
2746 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2747 void __user *buffer, size_t *lenp, loff_t *ppos)
2754 net = (struct net *)ctl->extra1;
2755 delay = net->ipv6.sysctl.flush_delay;
2756 proc_dointvec(ctl, write, buffer, lenp, ppos);
2757 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2761 ctl_table ipv6_route_table_template[] = {
2763 .procname = "flush",
2764 .data = &init_net.ipv6.sysctl.flush_delay,
2765 .maxlen = sizeof(int),
2767 .proc_handler = ipv6_sysctl_rtcache_flush
2770 .procname = "gc_thresh",
2771 .data = &ip6_dst_ops_template.gc_thresh,
2772 .maxlen = sizeof(int),
2774 .proc_handler = proc_dointvec,
2777 .procname = "max_size",
2778 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2779 .maxlen = sizeof(int),
2781 .proc_handler = proc_dointvec,
2784 .procname = "gc_min_interval",
2785 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2786 .maxlen = sizeof(int),
2788 .proc_handler = proc_dointvec_jiffies,
2791 .procname = "gc_timeout",
2792 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2793 .maxlen = sizeof(int),
2795 .proc_handler = proc_dointvec_jiffies,
2798 .procname = "gc_interval",
2799 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2800 .maxlen = sizeof(int),
2802 .proc_handler = proc_dointvec_jiffies,
2805 .procname = "gc_elasticity",
2806 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2807 .maxlen = sizeof(int),
2809 .proc_handler = proc_dointvec,
2812 .procname = "mtu_expires",
2813 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2814 .maxlen = sizeof(int),
2816 .proc_handler = proc_dointvec_jiffies,
2819 .procname = "min_adv_mss",
2820 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2821 .maxlen = sizeof(int),
2823 .proc_handler = proc_dointvec,
2826 .procname = "gc_min_interval_ms",
2827 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2828 .maxlen = sizeof(int),
2830 .proc_handler = proc_dointvec_ms_jiffies,
2835 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2837 struct ctl_table *table;
2839 table = kmemdup(ipv6_route_table_template,
2840 sizeof(ipv6_route_table_template),
2844 table[0].data = &net->ipv6.sysctl.flush_delay;
2845 table[0].extra1 = net;
2846 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2847 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2848 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2849 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2850 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2851 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2852 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2853 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2854 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2861 static int __net_init ip6_route_net_init(struct net *net)
2865 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2866 sizeof(net->ipv6.ip6_dst_ops));
2868 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2869 goto out_ip6_dst_ops;
2871 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2872 sizeof(*net->ipv6.ip6_null_entry),
2874 if (!net->ipv6.ip6_null_entry)
2875 goto out_ip6_dst_entries;
2876 net->ipv6.ip6_null_entry->dst.path =
2877 (struct dst_entry *)net->ipv6.ip6_null_entry;
2878 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2879 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2880 ip6_template_metrics, true);
2882 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2883 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2884 sizeof(*net->ipv6.ip6_prohibit_entry),
2886 if (!net->ipv6.ip6_prohibit_entry)
2887 goto out_ip6_null_entry;
2888 net->ipv6.ip6_prohibit_entry->dst.path =
2889 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2890 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2891 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2892 ip6_template_metrics, true);
2894 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2895 sizeof(*net->ipv6.ip6_blk_hole_entry),
2897 if (!net->ipv6.ip6_blk_hole_entry)
2898 goto out_ip6_prohibit_entry;
2899 net->ipv6.ip6_blk_hole_entry->dst.path =
2900 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2901 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2902 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2903 ip6_template_metrics, true);
2906 net->ipv6.sysctl.flush_delay = 0;
2907 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2908 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2909 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2910 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2911 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2912 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2913 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2915 #ifdef CONFIG_PROC_FS
2916 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2917 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2919 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2925 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2926 out_ip6_prohibit_entry:
2927 kfree(net->ipv6.ip6_prohibit_entry);
2929 kfree(net->ipv6.ip6_null_entry);
2931 out_ip6_dst_entries:
2932 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2937 static void __net_exit ip6_route_net_exit(struct net *net)
2939 #ifdef CONFIG_PROC_FS
2940 proc_net_remove(net, "ipv6_route");
2941 proc_net_remove(net, "rt6_stats");
2943 kfree(net->ipv6.ip6_null_entry);
2944 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2945 kfree(net->ipv6.ip6_prohibit_entry);
2946 kfree(net->ipv6.ip6_blk_hole_entry);
2948 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2951 static struct pernet_operations ip6_route_net_ops = {
2952 .init = ip6_route_net_init,
2953 .exit = ip6_route_net_exit,
2956 static struct notifier_block ip6_route_dev_notifier = {
2957 .notifier_call = ip6_route_dev_notify,
2961 int __init ip6_route_init(void)
2966 ip6_dst_ops_template.kmem_cachep =
2967 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2968 SLAB_HWCACHE_ALIGN, NULL);
2969 if (!ip6_dst_ops_template.kmem_cachep)
2972 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2974 goto out_kmem_cache;
2976 ret = register_pernet_subsys(&ip6_route_net_ops);
2978 goto out_dst_entries;
2980 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2982 /* Registering of the loopback is done before this portion of code,
2983 * the loopback reference in rt6_info will not be taken, do it
2984 * manually for init_net */
2985 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2986 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2987 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2988 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2989 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2990 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2991 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2995 goto out_register_subsys;
3001 ret = fib6_rules_init();
3006 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3007 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3008 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3009 goto fib6_rules_init;
3011 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3013 goto fib6_rules_init;
3019 fib6_rules_cleanup();
3024 out_register_subsys:
3025 unregister_pernet_subsys(&ip6_route_net_ops);
3027 dst_entries_destroy(&ip6_dst_blackhole_ops);
3029 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3033 void ip6_route_cleanup(void)
3035 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3036 fib6_rules_cleanup();
3039 unregister_pernet_subsys(&ip6_route_net_ops);
3040 dst_entries_destroy(&ip6_dst_blackhole_ops);
3041 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);