2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
61 #include <asm/uaccess.h>
64 #include <linux/sysctl.h>
67 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
68 const struct in6_addr *dest);
69 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
70 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
71 static unsigned int ip6_mtu(const struct dst_entry *dst);
72 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
73 static void ip6_dst_destroy(struct dst_entry *);
74 static void ip6_dst_ifdown(struct dst_entry *,
75 struct net_device *dev, int how);
76 static int ip6_dst_gc(struct dst_ops *ops);
78 static int ip6_pkt_discard(struct sk_buff *skb);
79 static int ip6_pkt_discard_out(struct sk_buff *skb);
80 static void ip6_link_failure(struct sk_buff *skb);
81 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
83 #ifdef CONFIG_IPV6_ROUTE_INFO
84 static struct rt6_info *rt6_add_route_info(struct net *net,
85 const struct in6_addr *prefix, int prefixlen,
86 const struct in6_addr *gwaddr, int ifindex,
88 static struct rt6_info *rt6_get_route_info(struct net *net,
89 const struct in6_addr *prefix, int prefixlen,
90 const struct in6_addr *gwaddr, int ifindex);
93 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
95 struct rt6_info *rt = (struct rt6_info *) dst;
96 struct inet_peer *peer;
99 if (!(rt->dst.flags & DST_HOST))
102 peer = rt6_get_peer_create(rt);
104 u32 *old_p = __DST_METRICS_PTR(old);
105 unsigned long prev, new;
108 if (inet_metrics_new(peer))
109 memcpy(p, old_p, sizeof(u32) * RTAX_MAX);
111 new = (unsigned long) p;
112 prev = cmpxchg(&dst->_metrics, old, new);
115 p = __DST_METRICS_PTR(prev);
116 if (prev & DST_METRICS_READ_ONLY)
123 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
127 struct in6_addr *p = &rt->rt6i_gateway;
129 if (!ipv6_addr_any(p))
130 return (const void *) p;
132 return &ipv6_hdr(skb)->daddr;
136 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
140 struct rt6_info *rt = (struct rt6_info *) dst;
143 daddr = choose_neigh_daddr(rt, skb, daddr);
144 n = __ipv6_neigh_lookup(&nd_tbl, dst->dev, daddr);
147 return neigh_create(&nd_tbl, daddr, dst->dev);
150 static int rt6_bind_neighbour(struct rt6_info *rt, struct net_device *dev)
152 struct neighbour *n = __ipv6_neigh_lookup(&nd_tbl, dev, &rt->rt6i_gateway);
154 n = neigh_create(&nd_tbl, &rt->rt6i_gateway, dev);
158 dst_set_neighbour(&rt->dst, n);
163 static struct dst_ops ip6_dst_ops_template = {
165 .protocol = cpu_to_be16(ETH_P_IPV6),
168 .check = ip6_dst_check,
169 .default_advmss = ip6_default_advmss,
171 .cow_metrics = ipv6_cow_metrics,
172 .destroy = ip6_dst_destroy,
173 .ifdown = ip6_dst_ifdown,
174 .negative_advice = ip6_negative_advice,
175 .link_failure = ip6_link_failure,
176 .update_pmtu = ip6_rt_update_pmtu,
177 .local_out = __ip6_local_out,
178 .neigh_lookup = ip6_neigh_lookup,
181 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
183 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
185 return mtu ? : dst->dev->mtu;
188 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
192 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
198 static struct dst_ops ip6_dst_blackhole_ops = {
200 .protocol = cpu_to_be16(ETH_P_IPV6),
201 .destroy = ip6_dst_destroy,
202 .check = ip6_dst_check,
203 .mtu = ip6_blackhole_mtu,
204 .default_advmss = ip6_default_advmss,
205 .update_pmtu = ip6_rt_blackhole_update_pmtu,
206 .cow_metrics = ip6_rt_blackhole_cow_metrics,
207 .neigh_lookup = ip6_neigh_lookup,
210 static const u32 ip6_template_metrics[RTAX_MAX] = {
211 [RTAX_HOPLIMIT - 1] = 255,
214 static struct rt6_info ip6_null_entry_template = {
216 .__refcnt = ATOMIC_INIT(1),
219 .error = -ENETUNREACH,
220 .input = ip6_pkt_discard,
221 .output = ip6_pkt_discard_out,
223 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
224 .rt6i_protocol = RTPROT_KERNEL,
225 .rt6i_metric = ~(u32) 0,
226 .rt6i_ref = ATOMIC_INIT(1),
229 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
231 static int ip6_pkt_prohibit(struct sk_buff *skb);
232 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
234 static struct rt6_info ip6_prohibit_entry_template = {
236 .__refcnt = ATOMIC_INIT(1),
240 .input = ip6_pkt_prohibit,
241 .output = ip6_pkt_prohibit_out,
243 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
244 .rt6i_protocol = RTPROT_KERNEL,
245 .rt6i_metric = ~(u32) 0,
246 .rt6i_ref = ATOMIC_INIT(1),
249 static struct rt6_info ip6_blk_hole_entry_template = {
251 .__refcnt = ATOMIC_INIT(1),
255 .input = dst_discard,
256 .output = dst_discard,
258 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
259 .rt6i_protocol = RTPROT_KERNEL,
260 .rt6i_metric = ~(u32) 0,
261 .rt6i_ref = ATOMIC_INIT(1),
266 /* allocate dst with ip6_dst_ops */
267 static inline struct rt6_info *ip6_dst_alloc(struct net *net,
268 struct net_device *dev,
270 struct fib6_table *table)
272 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
276 memset(&rt->rt6i_table, 0,
277 sizeof(*rt) - sizeof(struct dst_entry));
278 rt6_init_peer(rt, table ? &table->tb6_peers : net->ipv6.peers);
283 static void ip6_dst_destroy(struct dst_entry *dst)
285 struct rt6_info *rt = (struct rt6_info *)dst;
286 struct inet6_dev *idev = rt->rt6i_idev;
288 if (!(rt->dst.flags & DST_HOST))
289 dst_destroy_metrics_generic(dst);
292 rt->rt6i_idev = NULL;
296 if (!(rt->rt6i_flags & RTF_EXPIRES) && dst->from)
297 dst_release(dst->from);
299 if (rt6_has_peer(rt)) {
300 struct inet_peer *peer = rt6_peer_ptr(rt);
305 static atomic_t __rt6_peer_genid = ATOMIC_INIT(0);
307 static u32 rt6_peer_genid(void)
309 return atomic_read(&__rt6_peer_genid);
312 void rt6_bind_peer(struct rt6_info *rt, int create)
314 struct inet_peer_base *base;
315 struct inet_peer *peer;
317 base = inetpeer_base_ptr(rt->_rt6i_peer);
321 peer = inet_getpeer_v6(base, &rt->rt6i_dst.addr, create);
323 if (!rt6_set_peer(rt, peer))
326 rt->rt6i_peer_genid = rt6_peer_genid();
330 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
333 struct rt6_info *rt = (struct rt6_info *)dst;
334 struct inet6_dev *idev = rt->rt6i_idev;
335 struct net_device *loopback_dev =
336 dev_net(dev)->loopback_dev;
338 if (dev != loopback_dev && idev && idev->dev == dev) {
339 struct inet6_dev *loopback_idev =
340 in6_dev_get(loopback_dev);
342 rt->rt6i_idev = loopback_idev;
348 static bool rt6_check_expired(const struct rt6_info *rt)
350 struct rt6_info *ort = NULL;
352 if (rt->rt6i_flags & RTF_EXPIRES) {
353 if (time_after(jiffies, rt->dst.expires))
355 } else if (rt->dst.from) {
356 ort = (struct rt6_info *) rt->dst.from;
357 return (ort->rt6i_flags & RTF_EXPIRES) &&
358 time_after(jiffies, ort->dst.expires);
363 static bool rt6_need_strict(const struct in6_addr *daddr)
365 return ipv6_addr_type(daddr) &
366 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
370 * Route lookup. Any table->tb6_lock is implied.
373 static inline struct rt6_info *rt6_device_match(struct net *net,
375 const struct in6_addr *saddr,
379 struct rt6_info *local = NULL;
380 struct rt6_info *sprt;
382 if (!oif && ipv6_addr_any(saddr))
385 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
386 struct net_device *dev = sprt->dst.dev;
389 if (dev->ifindex == oif)
391 if (dev->flags & IFF_LOOPBACK) {
392 if (!sprt->rt6i_idev ||
393 sprt->rt6i_idev->dev->ifindex != oif) {
394 if (flags & RT6_LOOKUP_F_IFACE && oif)
396 if (local && (!oif ||
397 local->rt6i_idev->dev->ifindex == oif))
403 if (ipv6_chk_addr(net, saddr, dev,
404 flags & RT6_LOOKUP_F_IFACE))
413 if (flags & RT6_LOOKUP_F_IFACE)
414 return net->ipv6.ip6_null_entry;
420 #ifdef CONFIG_IPV6_ROUTER_PREF
421 static void rt6_probe(struct rt6_info *rt)
423 struct neighbour *neigh;
425 * Okay, this does not seem to be appropriate
426 * for now, however, we need to check if it
427 * is really so; aka Router Reachability Probing.
429 * Router Reachability Probe MUST be rate-limited
430 * to no more than one per minute.
433 neigh = rt ? dst_get_neighbour_noref(&rt->dst) : NULL;
434 if (!neigh || (neigh->nud_state & NUD_VALID))
436 read_lock_bh(&neigh->lock);
437 if (!(neigh->nud_state & NUD_VALID) &&
438 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
439 struct in6_addr mcaddr;
440 struct in6_addr *target;
442 neigh->updated = jiffies;
443 read_unlock_bh(&neigh->lock);
445 target = (struct in6_addr *)&neigh->primary_key;
446 addrconf_addr_solict_mult(target, &mcaddr);
447 ndisc_send_ns(rt->dst.dev, NULL, target, &mcaddr, NULL);
449 read_unlock_bh(&neigh->lock);
455 static inline void rt6_probe(struct rt6_info *rt)
461 * Default Router Selection (RFC 2461 6.3.6)
463 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
465 struct net_device *dev = rt->dst.dev;
466 if (!oif || dev->ifindex == oif)
468 if ((dev->flags & IFF_LOOPBACK) &&
469 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
474 static inline int rt6_check_neigh(struct rt6_info *rt)
476 struct neighbour *neigh;
480 neigh = dst_get_neighbour_noref(&rt->dst);
481 if (rt->rt6i_flags & RTF_NONEXTHOP ||
482 !(rt->rt6i_flags & RTF_GATEWAY))
485 read_lock_bh(&neigh->lock);
486 if (neigh->nud_state & NUD_VALID)
488 #ifdef CONFIG_IPV6_ROUTER_PREF
489 else if (neigh->nud_state & NUD_FAILED)
494 read_unlock_bh(&neigh->lock);
501 static int rt6_score_route(struct rt6_info *rt, int oif,
506 m = rt6_check_dev(rt, oif);
507 if (!m && (strict & RT6_LOOKUP_F_IFACE))
509 #ifdef CONFIG_IPV6_ROUTER_PREF
510 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
512 n = rt6_check_neigh(rt);
513 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
518 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
519 int *mpri, struct rt6_info *match)
523 if (rt6_check_expired(rt))
526 m = rt6_score_route(rt, oif, strict);
531 if (strict & RT6_LOOKUP_F_REACHABLE)
535 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
543 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
544 struct rt6_info *rr_head,
545 u32 metric, int oif, int strict)
547 struct rt6_info *rt, *match;
551 for (rt = rr_head; rt && rt->rt6i_metric == metric;
552 rt = rt->dst.rt6_next)
553 match = find_match(rt, oif, strict, &mpri, match);
554 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
555 rt = rt->dst.rt6_next)
556 match = find_match(rt, oif, strict, &mpri, match);
561 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
563 struct rt6_info *match, *rt0;
568 fn->rr_ptr = rt0 = fn->leaf;
570 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
573 (strict & RT6_LOOKUP_F_REACHABLE)) {
574 struct rt6_info *next = rt0->dst.rt6_next;
576 /* no entries matched; do round-robin */
577 if (!next || next->rt6i_metric != rt0->rt6i_metric)
584 net = dev_net(rt0->dst.dev);
585 return match ? match : net->ipv6.ip6_null_entry;
588 #ifdef CONFIG_IPV6_ROUTE_INFO
589 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
590 const struct in6_addr *gwaddr)
592 struct net *net = dev_net(dev);
593 struct route_info *rinfo = (struct route_info *) opt;
594 struct in6_addr prefix_buf, *prefix;
596 unsigned long lifetime;
599 if (len < sizeof(struct route_info)) {
603 /* Sanity check for prefix_len and length */
604 if (rinfo->length > 3) {
606 } else if (rinfo->prefix_len > 128) {
608 } else if (rinfo->prefix_len > 64) {
609 if (rinfo->length < 2) {
612 } else if (rinfo->prefix_len > 0) {
613 if (rinfo->length < 1) {
618 pref = rinfo->route_pref;
619 if (pref == ICMPV6_ROUTER_PREF_INVALID)
622 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
624 if (rinfo->length == 3)
625 prefix = (struct in6_addr *)rinfo->prefix;
627 /* this function is safe */
628 ipv6_addr_prefix(&prefix_buf,
629 (struct in6_addr *)rinfo->prefix,
631 prefix = &prefix_buf;
634 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
637 if (rt && !lifetime) {
643 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
646 rt->rt6i_flags = RTF_ROUTEINFO |
647 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
650 if (!addrconf_finite_timeout(lifetime))
651 rt6_clean_expires(rt);
653 rt6_set_expires(rt, jiffies + HZ * lifetime);
655 dst_release(&rt->dst);
661 #define BACKTRACK(__net, saddr) \
663 if (rt == __net->ipv6.ip6_null_entry) { \
664 struct fib6_node *pn; \
666 if (fn->fn_flags & RTN_TL_ROOT) \
669 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
670 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
673 if (fn->fn_flags & RTN_RTINFO) \
679 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
680 struct fib6_table *table,
681 struct flowi6 *fl6, int flags)
683 struct fib6_node *fn;
686 read_lock_bh(&table->tb6_lock);
687 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
690 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
691 BACKTRACK(net, &fl6->saddr);
693 dst_use(&rt->dst, jiffies);
694 read_unlock_bh(&table->tb6_lock);
699 struct dst_entry * ip6_route_lookup(struct net *net, struct flowi6 *fl6,
702 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
704 EXPORT_SYMBOL_GPL(ip6_route_lookup);
706 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
707 const struct in6_addr *saddr, int oif, int strict)
709 struct flowi6 fl6 = {
713 struct dst_entry *dst;
714 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
717 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
718 flags |= RT6_LOOKUP_F_HAS_SADDR;
721 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
723 return (struct rt6_info *) dst;
730 EXPORT_SYMBOL(rt6_lookup);
732 /* ip6_ins_rt is called with FREE table->tb6_lock.
733 It takes new route entry, the addition fails by any reason the
734 route is freed. In any case, if caller does not hold it, it may
738 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
741 struct fib6_table *table;
743 table = rt->rt6i_table;
744 write_lock_bh(&table->tb6_lock);
745 err = fib6_add(&table->tb6_root, rt, info);
746 write_unlock_bh(&table->tb6_lock);
751 int ip6_ins_rt(struct rt6_info *rt)
753 struct nl_info info = {
754 .nl_net = dev_net(rt->dst.dev),
756 return __ip6_ins_rt(rt, &info);
759 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort,
760 const struct in6_addr *daddr,
761 const struct in6_addr *saddr)
769 rt = ip6_rt_copy(ort, daddr);
772 int attempts = !in_softirq();
774 if (!(rt->rt6i_flags & RTF_GATEWAY)) {
775 if (ort->rt6i_dst.plen != 128 &&
776 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
777 rt->rt6i_flags |= RTF_ANYCAST;
778 rt->rt6i_gateway = *daddr;
781 rt->rt6i_flags |= RTF_CACHE;
783 #ifdef CONFIG_IPV6_SUBTREES
784 if (rt->rt6i_src.plen && saddr) {
785 rt->rt6i_src.addr = *saddr;
786 rt->rt6i_src.plen = 128;
791 if (rt6_bind_neighbour(rt, rt->dst.dev)) {
792 struct net *net = dev_net(rt->dst.dev);
793 int saved_rt_min_interval =
794 net->ipv6.sysctl.ip6_rt_gc_min_interval;
795 int saved_rt_elasticity =
796 net->ipv6.sysctl.ip6_rt_gc_elasticity;
798 if (attempts-- > 0) {
799 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
800 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
802 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
804 net->ipv6.sysctl.ip6_rt_gc_elasticity =
806 net->ipv6.sysctl.ip6_rt_gc_min_interval =
807 saved_rt_min_interval;
811 net_warn_ratelimited("Neighbour table overflow\n");
820 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort,
821 const struct in6_addr *daddr)
823 struct rt6_info *rt = ip6_rt_copy(ort, daddr);
826 rt->rt6i_flags |= RTF_CACHE;
827 dst_set_neighbour(&rt->dst, neigh_clone(dst_get_neighbour_noref_raw(&ort->dst)));
832 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
833 struct flowi6 *fl6, int flags)
835 struct fib6_node *fn;
836 struct rt6_info *rt, *nrt;
840 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
842 strict |= flags & RT6_LOOKUP_F_IFACE;
845 read_lock_bh(&table->tb6_lock);
848 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
851 rt = rt6_select(fn, oif, strict | reachable);
853 BACKTRACK(net, &fl6->saddr);
854 if (rt == net->ipv6.ip6_null_entry ||
855 rt->rt6i_flags & RTF_CACHE)
859 read_unlock_bh(&table->tb6_lock);
861 if (!dst_get_neighbour_noref_raw(&rt->dst) && !(rt->rt6i_flags & RTF_NONEXTHOP))
862 nrt = rt6_alloc_cow(rt, &fl6->daddr, &fl6->saddr);
863 else if (!(rt->dst.flags & DST_HOST))
864 nrt = rt6_alloc_clone(rt, &fl6->daddr);
868 dst_release(&rt->dst);
869 rt = nrt ? : net->ipv6.ip6_null_entry;
873 err = ip6_ins_rt(nrt);
882 * Race condition! In the gap, when table->tb6_lock was
883 * released someone could insert this route. Relookup.
885 dst_release(&rt->dst);
894 read_unlock_bh(&table->tb6_lock);
896 rt->dst.lastuse = jiffies;
902 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
903 struct flowi6 *fl6, int flags)
905 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
908 static struct dst_entry *ip6_route_input_lookup(struct net *net,
909 struct net_device *dev,
910 struct flowi6 *fl6, int flags)
912 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
913 flags |= RT6_LOOKUP_F_IFACE;
915 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
918 void ip6_route_input(struct sk_buff *skb)
920 const struct ipv6hdr *iph = ipv6_hdr(skb);
921 struct net *net = dev_net(skb->dev);
922 int flags = RT6_LOOKUP_F_HAS_SADDR;
923 struct flowi6 fl6 = {
924 .flowi6_iif = skb->dev->ifindex,
927 .flowlabel = (* (__be32 *) iph) & IPV6_FLOWINFO_MASK,
928 .flowi6_mark = skb->mark,
929 .flowi6_proto = iph->nexthdr,
932 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
935 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
936 struct flowi6 *fl6, int flags)
938 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
941 struct dst_entry * ip6_route_output(struct net *net, const struct sock *sk,
946 fl6->flowi6_iif = net->loopback_dev->ifindex;
948 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
949 flags |= RT6_LOOKUP_F_IFACE;
951 if (!ipv6_addr_any(&fl6->saddr))
952 flags |= RT6_LOOKUP_F_HAS_SADDR;
954 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
956 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
959 EXPORT_SYMBOL(ip6_route_output);
961 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
963 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
964 struct dst_entry *new = NULL;
966 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, 0, 0);
968 memset(&rt->rt6i_table, 0, sizeof(*rt) - sizeof(struct dst_entry));
969 rt6_init_peer(rt, net->ipv6.peers);
974 new->input = dst_discard;
975 new->output = dst_discard;
977 if (dst_metrics_read_only(&ort->dst))
978 new->_metrics = ort->dst._metrics;
980 dst_copy_metrics(new, &ort->dst);
981 rt->rt6i_idev = ort->rt6i_idev;
983 in6_dev_hold(rt->rt6i_idev);
985 rt->rt6i_gateway = ort->rt6i_gateway;
986 rt->rt6i_flags = ort->rt6i_flags;
987 rt6_clean_expires(rt);
990 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
991 #ifdef CONFIG_IPV6_SUBTREES
992 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
998 dst_release(dst_orig);
999 return new ? new : ERR_PTR(-ENOMEM);
1003 * Destination cache support functions
1006 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1008 struct rt6_info *rt;
1010 rt = (struct rt6_info *) dst;
1012 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie)) {
1013 if (rt->rt6i_peer_genid != rt6_peer_genid()) {
1014 if (!rt6_has_peer(rt))
1015 rt6_bind_peer(rt, 0);
1016 rt->rt6i_peer_genid = rt6_peer_genid();
1023 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1025 struct rt6_info *rt = (struct rt6_info *) dst;
1028 if (rt->rt6i_flags & RTF_CACHE) {
1029 if (rt6_check_expired(rt)) {
1041 static void ip6_link_failure(struct sk_buff *skb)
1043 struct rt6_info *rt;
1045 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1047 rt = (struct rt6_info *) skb_dst(skb);
1049 if (rt->rt6i_flags & RTF_CACHE)
1050 rt6_update_expires(rt, 0);
1051 else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
1052 rt->rt6i_node->fn_sernum = -1;
1056 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
1058 struct rt6_info *rt6 = (struct rt6_info*)dst;
1061 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
1062 struct net *net = dev_net(dst->dev);
1064 rt6->rt6i_flags |= RTF_MODIFIED;
1065 if (mtu < IPV6_MIN_MTU) {
1066 u32 features = dst_metric(dst, RTAX_FEATURES);
1068 features |= RTAX_FEATURE_ALLFRAG;
1069 dst_metric_set(dst, RTAX_FEATURES, features);
1071 dst_metric_set(dst, RTAX_MTU, mtu);
1072 rt6_update_expires(rt6, net->ipv6.sysctl.ip6_rt_mtu_expires);
1076 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1079 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1080 struct dst_entry *dst;
1083 memset(&fl6, 0, sizeof(fl6));
1084 fl6.flowi6_oif = oif;
1085 fl6.flowi6_mark = mark;
1086 fl6.flowi6_flags = FLOWI_FLAG_PRECOW_METRICS;
1087 fl6.daddr = iph->daddr;
1088 fl6.saddr = iph->saddr;
1089 fl6.flowlabel = (*(__be32 *) iph) & IPV6_FLOWINFO_MASK;
1091 dst = ip6_route_output(net, NULL, &fl6);
1093 ip6_rt_update_pmtu(dst, ntohl(mtu));
1096 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1098 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1100 ip6_update_pmtu(skb, sock_net(sk), mtu,
1101 sk->sk_bound_dev_if, sk->sk_mark);
1103 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1105 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1107 struct net_device *dev = dst->dev;
1108 unsigned int mtu = dst_mtu(dst);
1109 struct net *net = dev_net(dev);
1111 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1113 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1114 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1117 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1118 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1119 * IPV6_MAXPLEN is also valid and means: "any MSS,
1120 * rely only on pmtu discovery"
1122 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1127 static unsigned int ip6_mtu(const struct dst_entry *dst)
1129 struct inet6_dev *idev;
1130 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
1138 idev = __in6_dev_get(dst->dev);
1140 mtu = idev->cnf.mtu6;
1146 static struct dst_entry *icmp6_dst_gc_list;
1147 static DEFINE_SPINLOCK(icmp6_dst_lock);
1149 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1150 struct neighbour *neigh,
1153 struct dst_entry *dst;
1154 struct rt6_info *rt;
1155 struct inet6_dev *idev = in6_dev_get(dev);
1156 struct net *net = dev_net(dev);
1158 if (unlikely(!idev))
1159 return ERR_PTR(-ENODEV);
1161 rt = ip6_dst_alloc(net, dev, 0, NULL);
1162 if (unlikely(!rt)) {
1164 dst = ERR_PTR(-ENOMEM);
1171 neigh = ip6_neigh_lookup(&rt->dst, NULL, &fl6->daddr);
1172 if (IS_ERR(neigh)) {
1175 return ERR_CAST(neigh);
1179 rt->dst.flags |= DST_HOST;
1180 rt->dst.output = ip6_output;
1181 dst_set_neighbour(&rt->dst, neigh);
1182 atomic_set(&rt->dst.__refcnt, 1);
1183 rt->rt6i_dst.addr = fl6->daddr;
1184 rt->rt6i_dst.plen = 128;
1185 rt->rt6i_idev = idev;
1186 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1188 spin_lock_bh(&icmp6_dst_lock);
1189 rt->dst.next = icmp6_dst_gc_list;
1190 icmp6_dst_gc_list = &rt->dst;
1191 spin_unlock_bh(&icmp6_dst_lock);
1193 fib6_force_start_gc(net);
1195 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1201 int icmp6_dst_gc(void)
1203 struct dst_entry *dst, **pprev;
1206 spin_lock_bh(&icmp6_dst_lock);
1207 pprev = &icmp6_dst_gc_list;
1209 while ((dst = *pprev) != NULL) {
1210 if (!atomic_read(&dst->__refcnt)) {
1219 spin_unlock_bh(&icmp6_dst_lock);
1224 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1227 struct dst_entry *dst, **pprev;
1229 spin_lock_bh(&icmp6_dst_lock);
1230 pprev = &icmp6_dst_gc_list;
1231 while ((dst = *pprev) != NULL) {
1232 struct rt6_info *rt = (struct rt6_info *) dst;
1233 if (func(rt, arg)) {
1240 spin_unlock_bh(&icmp6_dst_lock);
1243 static int ip6_dst_gc(struct dst_ops *ops)
1245 unsigned long now = jiffies;
1246 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1247 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1248 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1249 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1250 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1251 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1254 entries = dst_entries_get_fast(ops);
1255 if (time_after(rt_last_gc + rt_min_interval, now) &&
1256 entries <= rt_max_size)
1259 net->ipv6.ip6_rt_gc_expire++;
1260 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1261 net->ipv6.ip6_rt_last_gc = now;
1262 entries = dst_entries_get_slow(ops);
1263 if (entries < ops->gc_thresh)
1264 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1266 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1267 return entries > rt_max_size;
1270 /* Clean host part of a prefix. Not necessary in radix tree,
1271 but results in cleaner routing tables.
1273 Remove it only when all the things will work!
1276 int ip6_dst_hoplimit(struct dst_entry *dst)
1278 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1279 if (hoplimit == 0) {
1280 struct net_device *dev = dst->dev;
1281 struct inet6_dev *idev;
1284 idev = __in6_dev_get(dev);
1286 hoplimit = idev->cnf.hop_limit;
1288 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1293 EXPORT_SYMBOL(ip6_dst_hoplimit);
1299 int ip6_route_add(struct fib6_config *cfg)
1302 struct net *net = cfg->fc_nlinfo.nl_net;
1303 struct rt6_info *rt = NULL;
1304 struct net_device *dev = NULL;
1305 struct inet6_dev *idev = NULL;
1306 struct fib6_table *table;
1309 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1311 #ifndef CONFIG_IPV6_SUBTREES
1312 if (cfg->fc_src_len)
1315 if (cfg->fc_ifindex) {
1317 dev = dev_get_by_index(net, cfg->fc_ifindex);
1320 idev = in6_dev_get(dev);
1325 if (cfg->fc_metric == 0)
1326 cfg->fc_metric = IP6_RT_PRIO_USER;
1329 if (cfg->fc_nlinfo.nlh &&
1330 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1331 table = fib6_get_table(net, cfg->fc_table);
1333 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1334 table = fib6_new_table(net, cfg->fc_table);
1337 table = fib6_new_table(net, cfg->fc_table);
1343 rt = ip6_dst_alloc(net, NULL, DST_NOCOUNT, table);
1350 rt->dst.obsolete = -1;
1352 if (cfg->fc_flags & RTF_EXPIRES)
1353 rt6_set_expires(rt, jiffies +
1354 clock_t_to_jiffies(cfg->fc_expires));
1356 rt6_clean_expires(rt);
1358 if (cfg->fc_protocol == RTPROT_UNSPEC)
1359 cfg->fc_protocol = RTPROT_BOOT;
1360 rt->rt6i_protocol = cfg->fc_protocol;
1362 addr_type = ipv6_addr_type(&cfg->fc_dst);
1364 if (addr_type & IPV6_ADDR_MULTICAST)
1365 rt->dst.input = ip6_mc_input;
1366 else if (cfg->fc_flags & RTF_LOCAL)
1367 rt->dst.input = ip6_input;
1369 rt->dst.input = ip6_forward;
1371 rt->dst.output = ip6_output;
1373 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1374 rt->rt6i_dst.plen = cfg->fc_dst_len;
1375 if (rt->rt6i_dst.plen == 128)
1376 rt->dst.flags |= DST_HOST;
1378 if (!(rt->dst.flags & DST_HOST) && cfg->fc_mx) {
1379 u32 *metrics = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1384 dst_init_metrics(&rt->dst, metrics, 0);
1386 #ifdef CONFIG_IPV6_SUBTREES
1387 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1388 rt->rt6i_src.plen = cfg->fc_src_len;
1391 rt->rt6i_metric = cfg->fc_metric;
1393 /* We cannot add true routes via loopback here,
1394 they would result in kernel looping; promote them to reject routes
1396 if ((cfg->fc_flags & RTF_REJECT) ||
1397 (dev && (dev->flags & IFF_LOOPBACK) &&
1398 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1399 !(cfg->fc_flags & RTF_LOCAL))) {
1400 /* hold loopback dev/idev if we haven't done so. */
1401 if (dev != net->loopback_dev) {
1406 dev = net->loopback_dev;
1408 idev = in6_dev_get(dev);
1414 rt->dst.output = ip6_pkt_discard_out;
1415 rt->dst.input = ip6_pkt_discard;
1416 rt->dst.error = -ENETUNREACH;
1417 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1421 if (cfg->fc_flags & RTF_GATEWAY) {
1422 const struct in6_addr *gw_addr;
1425 gw_addr = &cfg->fc_gateway;
1426 rt->rt6i_gateway = *gw_addr;
1427 gwa_type = ipv6_addr_type(gw_addr);
1429 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1430 struct rt6_info *grt;
1432 /* IPv6 strictly inhibits using not link-local
1433 addresses as nexthop address.
1434 Otherwise, router will not able to send redirects.
1435 It is very good, but in some (rare!) circumstances
1436 (SIT, PtP, NBMA NOARP links) it is handy to allow
1437 some exceptions. --ANK
1440 if (!(gwa_type & IPV6_ADDR_UNICAST))
1443 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1445 err = -EHOSTUNREACH;
1449 if (dev != grt->dst.dev) {
1450 dst_release(&grt->dst);
1455 idev = grt->rt6i_idev;
1457 in6_dev_hold(grt->rt6i_idev);
1459 if (!(grt->rt6i_flags & RTF_GATEWAY))
1461 dst_release(&grt->dst);
1467 if (!dev || (dev->flags & IFF_LOOPBACK))
1475 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1476 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1480 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1481 rt->rt6i_prefsrc.plen = 128;
1483 rt->rt6i_prefsrc.plen = 0;
1485 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1486 err = rt6_bind_neighbour(rt, dev);
1491 rt->rt6i_flags = cfg->fc_flags;
1498 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1499 int type = nla_type(nla);
1502 if (type > RTAX_MAX) {
1507 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1513 rt->rt6i_idev = idev;
1514 rt->rt6i_table = table;
1516 cfg->fc_nlinfo.nl_net = dev_net(dev);
1518 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1530 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1533 struct fib6_table *table;
1534 struct net *net = dev_net(rt->dst.dev);
1536 if (rt == net->ipv6.ip6_null_entry)
1539 table = rt->rt6i_table;
1540 write_lock_bh(&table->tb6_lock);
1542 err = fib6_del(rt, info);
1543 dst_release(&rt->dst);
1545 write_unlock_bh(&table->tb6_lock);
1550 int ip6_del_rt(struct rt6_info *rt)
1552 struct nl_info info = {
1553 .nl_net = dev_net(rt->dst.dev),
1555 return __ip6_del_rt(rt, &info);
1558 static int ip6_route_del(struct fib6_config *cfg)
1560 struct fib6_table *table;
1561 struct fib6_node *fn;
1562 struct rt6_info *rt;
1565 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1569 read_lock_bh(&table->tb6_lock);
1571 fn = fib6_locate(&table->tb6_root,
1572 &cfg->fc_dst, cfg->fc_dst_len,
1573 &cfg->fc_src, cfg->fc_src_len);
1576 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1577 if (cfg->fc_ifindex &&
1579 rt->dst.dev->ifindex != cfg->fc_ifindex))
1581 if (cfg->fc_flags & RTF_GATEWAY &&
1582 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1584 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1587 read_unlock_bh(&table->tb6_lock);
1589 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1592 read_unlock_bh(&table->tb6_lock);
1600 struct ip6rd_flowi {
1602 struct in6_addr gateway;
1605 static struct rt6_info *__ip6_route_redirect(struct net *net,
1606 struct fib6_table *table,
1610 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1611 struct rt6_info *rt;
1612 struct fib6_node *fn;
1615 * Get the "current" route for this destination and
1616 * check if the redirect has come from approriate router.
1618 * RFC 2461 specifies that redirects should only be
1619 * accepted if they come from the nexthop to the target.
1620 * Due to the way the routes are chosen, this notion
1621 * is a bit fuzzy and one might need to check all possible
1625 read_lock_bh(&table->tb6_lock);
1626 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1628 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1630 * Current route is on-link; redirect is always invalid.
1632 * Seems, previous statement is not true. It could
1633 * be node, which looks for us as on-link (f.e. proxy ndisc)
1634 * But then router serving it might decide, that we should
1635 * know truth 8)8) --ANK (980726).
1637 if (rt6_check_expired(rt))
1639 if (!(rt->rt6i_flags & RTF_GATEWAY))
1641 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1643 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1649 rt = net->ipv6.ip6_null_entry;
1650 BACKTRACK(net, &fl6->saddr);
1654 read_unlock_bh(&table->tb6_lock);
1659 static struct rt6_info *ip6_route_redirect(const struct in6_addr *dest,
1660 const struct in6_addr *src,
1661 const struct in6_addr *gateway,
1662 struct net_device *dev)
1664 int flags = RT6_LOOKUP_F_HAS_SADDR;
1665 struct net *net = dev_net(dev);
1666 struct ip6rd_flowi rdfl = {
1668 .flowi6_oif = dev->ifindex,
1674 rdfl.gateway = *gateway;
1676 if (rt6_need_strict(dest))
1677 flags |= RT6_LOOKUP_F_IFACE;
1679 return (struct rt6_info *)fib6_rule_lookup(net, &rdfl.fl6,
1680 flags, __ip6_route_redirect);
1683 void rt6_redirect(const struct in6_addr *dest, const struct in6_addr *src,
1684 const struct in6_addr *saddr,
1685 struct neighbour *neigh, u8 *lladdr, int on_link)
1687 struct rt6_info *rt, *nrt = NULL;
1688 struct netevent_redirect netevent;
1689 struct net *net = dev_net(neigh->dev);
1690 struct neighbour *old_neigh;
1692 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1694 if (rt == net->ipv6.ip6_null_entry) {
1695 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1700 * We have finally decided to accept it.
1703 neigh_update(neigh, lladdr, NUD_STALE,
1704 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1705 NEIGH_UPDATE_F_OVERRIDE|
1706 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1707 NEIGH_UPDATE_F_ISROUTER))
1711 * Redirect received -> path was valid.
1712 * Look, redirects are sent only in response to data packets,
1713 * so that this nexthop apparently is reachable. --ANK
1715 dst_confirm(&rt->dst);
1717 /* Duplicate redirect: silently ignore. */
1718 old_neigh = dst_get_neighbour_noref_raw(&rt->dst);
1719 if (neigh == old_neigh)
1722 nrt = ip6_rt_copy(rt, dest);
1726 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1728 nrt->rt6i_flags &= ~RTF_GATEWAY;
1730 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1731 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1733 if (ip6_ins_rt(nrt))
1736 netevent.old = &rt->dst;
1737 netevent.old_neigh = old_neigh;
1738 netevent.new = &nrt->dst;
1739 netevent.new_neigh = neigh;
1740 netevent.daddr = dest;
1741 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1743 if (rt->rt6i_flags & RTF_CACHE) {
1749 dst_release(&rt->dst);
1753 * Misc support functions
1756 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1757 const struct in6_addr *dest)
1759 struct net *net = dev_net(ort->dst.dev);
1760 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1764 rt->dst.input = ort->dst.input;
1765 rt->dst.output = ort->dst.output;
1766 rt->dst.flags |= DST_HOST;
1768 rt->rt6i_dst.addr = *dest;
1769 rt->rt6i_dst.plen = 128;
1770 dst_copy_metrics(&rt->dst, &ort->dst);
1771 rt->dst.error = ort->dst.error;
1772 rt->rt6i_idev = ort->rt6i_idev;
1774 in6_dev_hold(rt->rt6i_idev);
1775 rt->dst.lastuse = jiffies;
1777 rt->rt6i_gateway = ort->rt6i_gateway;
1778 rt->rt6i_flags = ort->rt6i_flags;
1779 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1780 (RTF_DEFAULT | RTF_ADDRCONF))
1781 rt6_set_from(rt, ort);
1783 rt6_clean_expires(rt);
1784 rt->rt6i_metric = 0;
1786 #ifdef CONFIG_IPV6_SUBTREES
1787 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1789 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1790 rt->rt6i_table = ort->rt6i_table;
1795 #ifdef CONFIG_IPV6_ROUTE_INFO
1796 static struct rt6_info *rt6_get_route_info(struct net *net,
1797 const struct in6_addr *prefix, int prefixlen,
1798 const struct in6_addr *gwaddr, int ifindex)
1800 struct fib6_node *fn;
1801 struct rt6_info *rt = NULL;
1802 struct fib6_table *table;
1804 table = fib6_get_table(net, RT6_TABLE_INFO);
1808 write_lock_bh(&table->tb6_lock);
1809 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1813 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1814 if (rt->dst.dev->ifindex != ifindex)
1816 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1818 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1824 write_unlock_bh(&table->tb6_lock);
1828 static struct rt6_info *rt6_add_route_info(struct net *net,
1829 const struct in6_addr *prefix, int prefixlen,
1830 const struct in6_addr *gwaddr, int ifindex,
1833 struct fib6_config cfg = {
1834 .fc_table = RT6_TABLE_INFO,
1835 .fc_metric = IP6_RT_PRIO_USER,
1836 .fc_ifindex = ifindex,
1837 .fc_dst_len = prefixlen,
1838 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1839 RTF_UP | RTF_PREF(pref),
1841 .fc_nlinfo.nlh = NULL,
1842 .fc_nlinfo.nl_net = net,
1845 cfg.fc_dst = *prefix;
1846 cfg.fc_gateway = *gwaddr;
1848 /* We should treat it as a default route if prefix length is 0. */
1850 cfg.fc_flags |= RTF_DEFAULT;
1852 ip6_route_add(&cfg);
1854 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1858 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1860 struct rt6_info *rt;
1861 struct fib6_table *table;
1863 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1867 write_lock_bh(&table->tb6_lock);
1868 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1869 if (dev == rt->dst.dev &&
1870 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1871 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1876 write_unlock_bh(&table->tb6_lock);
1880 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1881 struct net_device *dev,
1884 struct fib6_config cfg = {
1885 .fc_table = RT6_TABLE_DFLT,
1886 .fc_metric = IP6_RT_PRIO_USER,
1887 .fc_ifindex = dev->ifindex,
1888 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1889 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1891 .fc_nlinfo.nlh = NULL,
1892 .fc_nlinfo.nl_net = dev_net(dev),
1895 cfg.fc_gateway = *gwaddr;
1897 ip6_route_add(&cfg);
1899 return rt6_get_dflt_router(gwaddr, dev);
1902 void rt6_purge_dflt_routers(struct net *net)
1904 struct rt6_info *rt;
1905 struct fib6_table *table;
1907 /* NOTE: Keep consistent with rt6_get_dflt_router */
1908 table = fib6_get_table(net, RT6_TABLE_DFLT);
1913 read_lock_bh(&table->tb6_lock);
1914 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1915 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1917 read_unlock_bh(&table->tb6_lock);
1922 read_unlock_bh(&table->tb6_lock);
1925 static void rtmsg_to_fib6_config(struct net *net,
1926 struct in6_rtmsg *rtmsg,
1927 struct fib6_config *cfg)
1929 memset(cfg, 0, sizeof(*cfg));
1931 cfg->fc_table = RT6_TABLE_MAIN;
1932 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1933 cfg->fc_metric = rtmsg->rtmsg_metric;
1934 cfg->fc_expires = rtmsg->rtmsg_info;
1935 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1936 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1937 cfg->fc_flags = rtmsg->rtmsg_flags;
1939 cfg->fc_nlinfo.nl_net = net;
1941 cfg->fc_dst = rtmsg->rtmsg_dst;
1942 cfg->fc_src = rtmsg->rtmsg_src;
1943 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1946 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1948 struct fib6_config cfg;
1949 struct in6_rtmsg rtmsg;
1953 case SIOCADDRT: /* Add a route */
1954 case SIOCDELRT: /* Delete a route */
1955 if (!capable(CAP_NET_ADMIN))
1957 err = copy_from_user(&rtmsg, arg,
1958 sizeof(struct in6_rtmsg));
1962 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1967 err = ip6_route_add(&cfg);
1970 err = ip6_route_del(&cfg);
1984 * Drop the packet on the floor
1987 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1990 struct dst_entry *dst = skb_dst(skb);
1991 switch (ipstats_mib_noroutes) {
1992 case IPSTATS_MIB_INNOROUTES:
1993 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1994 if (type == IPV6_ADDR_ANY) {
1995 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1996 IPSTATS_MIB_INADDRERRORS);
2000 case IPSTATS_MIB_OUTNOROUTES:
2001 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2002 ipstats_mib_noroutes);
2005 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2010 static int ip6_pkt_discard(struct sk_buff *skb)
2012 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2015 static int ip6_pkt_discard_out(struct sk_buff *skb)
2017 skb->dev = skb_dst(skb)->dev;
2018 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2021 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2023 static int ip6_pkt_prohibit(struct sk_buff *skb)
2025 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2028 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2030 skb->dev = skb_dst(skb)->dev;
2031 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2037 * Allocate a dst for local (unicast / anycast) address.
2040 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2041 const struct in6_addr *addr,
2044 struct net *net = dev_net(idev->dev);
2045 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2049 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2050 return ERR_PTR(-ENOMEM);
2055 rt->dst.flags |= DST_HOST;
2056 rt->dst.input = ip6_input;
2057 rt->dst.output = ip6_output;
2058 rt->rt6i_idev = idev;
2059 rt->dst.obsolete = -1;
2061 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2063 rt->rt6i_flags |= RTF_ANYCAST;
2065 rt->rt6i_flags |= RTF_LOCAL;
2066 err = rt6_bind_neighbour(rt, rt->dst.dev);
2069 return ERR_PTR(err);
2072 rt->rt6i_dst.addr = *addr;
2073 rt->rt6i_dst.plen = 128;
2074 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2076 atomic_set(&rt->dst.__refcnt, 1);
2081 int ip6_route_get_saddr(struct net *net,
2082 struct rt6_info *rt,
2083 const struct in6_addr *daddr,
2085 struct in6_addr *saddr)
2087 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2089 if (rt->rt6i_prefsrc.plen)
2090 *saddr = rt->rt6i_prefsrc.addr;
2092 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2093 daddr, prefs, saddr);
2097 /* remove deleted ip from prefsrc entries */
2098 struct arg_dev_net_ip {
2099 struct net_device *dev;
2101 struct in6_addr *addr;
2104 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2106 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2107 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2108 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2110 if (((void *)rt->dst.dev == dev || !dev) &&
2111 rt != net->ipv6.ip6_null_entry &&
2112 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2113 /* remove prefsrc entry */
2114 rt->rt6i_prefsrc.plen = 0;
2119 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2121 struct net *net = dev_net(ifp->idev->dev);
2122 struct arg_dev_net_ip adni = {
2123 .dev = ifp->idev->dev,
2127 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2130 struct arg_dev_net {
2131 struct net_device *dev;
2135 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2137 const struct arg_dev_net *adn = arg;
2138 const struct net_device *dev = adn->dev;
2140 if ((rt->dst.dev == dev || !dev) &&
2141 rt != adn->net->ipv6.ip6_null_entry)
2147 void rt6_ifdown(struct net *net, struct net_device *dev)
2149 struct arg_dev_net adn = {
2154 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2155 icmp6_clean_all(fib6_ifdown, &adn);
2158 struct rt6_mtu_change_arg {
2159 struct net_device *dev;
2163 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2165 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2166 struct inet6_dev *idev;
2168 /* In IPv6 pmtu discovery is not optional,
2169 so that RTAX_MTU lock cannot disable it.
2170 We still use this lock to block changes
2171 caused by addrconf/ndisc.
2174 idev = __in6_dev_get(arg->dev);
2178 /* For administrative MTU increase, there is no way to discover
2179 IPv6 PMTU increase, so PMTU increase should be updated here.
2180 Since RFC 1981 doesn't include administrative MTU increase
2181 update PMTU increase is a MUST. (i.e. jumbo frame)
2184 If new MTU is less than route PMTU, this new MTU will be the
2185 lowest MTU in the path, update the route PMTU to reflect PMTU
2186 decreases; if new MTU is greater than route PMTU, and the
2187 old MTU is the lowest MTU in the path, update the route PMTU
2188 to reflect the increase. In this case if the other nodes' MTU
2189 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2192 if (rt->dst.dev == arg->dev &&
2193 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2194 (dst_mtu(&rt->dst) >= arg->mtu ||
2195 (dst_mtu(&rt->dst) < arg->mtu &&
2196 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2197 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2202 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2204 struct rt6_mtu_change_arg arg = {
2209 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2212 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2213 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2214 [RTA_OIF] = { .type = NLA_U32 },
2215 [RTA_IIF] = { .type = NLA_U32 },
2216 [RTA_PRIORITY] = { .type = NLA_U32 },
2217 [RTA_METRICS] = { .type = NLA_NESTED },
2220 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2221 struct fib6_config *cfg)
2224 struct nlattr *tb[RTA_MAX+1];
2227 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2232 rtm = nlmsg_data(nlh);
2233 memset(cfg, 0, sizeof(*cfg));
2235 cfg->fc_table = rtm->rtm_table;
2236 cfg->fc_dst_len = rtm->rtm_dst_len;
2237 cfg->fc_src_len = rtm->rtm_src_len;
2238 cfg->fc_flags = RTF_UP;
2239 cfg->fc_protocol = rtm->rtm_protocol;
2241 if (rtm->rtm_type == RTN_UNREACHABLE)
2242 cfg->fc_flags |= RTF_REJECT;
2244 if (rtm->rtm_type == RTN_LOCAL)
2245 cfg->fc_flags |= RTF_LOCAL;
2247 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2248 cfg->fc_nlinfo.nlh = nlh;
2249 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2251 if (tb[RTA_GATEWAY]) {
2252 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2253 cfg->fc_flags |= RTF_GATEWAY;
2257 int plen = (rtm->rtm_dst_len + 7) >> 3;
2259 if (nla_len(tb[RTA_DST]) < plen)
2262 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2266 int plen = (rtm->rtm_src_len + 7) >> 3;
2268 if (nla_len(tb[RTA_SRC]) < plen)
2271 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2274 if (tb[RTA_PREFSRC])
2275 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2278 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2280 if (tb[RTA_PRIORITY])
2281 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2283 if (tb[RTA_METRICS]) {
2284 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2285 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2289 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2296 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2298 struct fib6_config cfg;
2301 err = rtm_to_fib6_config(skb, nlh, &cfg);
2305 return ip6_route_del(&cfg);
2308 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2310 struct fib6_config cfg;
2313 err = rtm_to_fib6_config(skb, nlh, &cfg);
2317 return ip6_route_add(&cfg);
2320 static inline size_t rt6_nlmsg_size(void)
2322 return NLMSG_ALIGN(sizeof(struct rtmsg))
2323 + nla_total_size(16) /* RTA_SRC */
2324 + nla_total_size(16) /* RTA_DST */
2325 + nla_total_size(16) /* RTA_GATEWAY */
2326 + nla_total_size(16) /* RTA_PREFSRC */
2327 + nla_total_size(4) /* RTA_TABLE */
2328 + nla_total_size(4) /* RTA_IIF */
2329 + nla_total_size(4) /* RTA_OIF */
2330 + nla_total_size(4) /* RTA_PRIORITY */
2331 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2332 + nla_total_size(sizeof(struct rta_cacheinfo));
2335 static int rt6_fill_node(struct net *net,
2336 struct sk_buff *skb, struct rt6_info *rt,
2337 struct in6_addr *dst, struct in6_addr *src,
2338 int iif, int type, u32 pid, u32 seq,
2339 int prefix, int nowait, unsigned int flags)
2341 const struct inet_peer *peer;
2343 struct nlmsghdr *nlh;
2346 struct neighbour *n;
2349 if (prefix) { /* user wants prefix routes only */
2350 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2351 /* success since this is not a prefix route */
2356 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2360 rtm = nlmsg_data(nlh);
2361 rtm->rtm_family = AF_INET6;
2362 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2363 rtm->rtm_src_len = rt->rt6i_src.plen;
2366 table = rt->rt6i_table->tb6_id;
2368 table = RT6_TABLE_UNSPEC;
2369 rtm->rtm_table = table;
2370 if (nla_put_u32(skb, RTA_TABLE, table))
2371 goto nla_put_failure;
2372 if (rt->rt6i_flags & RTF_REJECT)
2373 rtm->rtm_type = RTN_UNREACHABLE;
2374 else if (rt->rt6i_flags & RTF_LOCAL)
2375 rtm->rtm_type = RTN_LOCAL;
2376 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2377 rtm->rtm_type = RTN_LOCAL;
2379 rtm->rtm_type = RTN_UNICAST;
2381 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2382 rtm->rtm_protocol = rt->rt6i_protocol;
2383 if (rt->rt6i_flags & RTF_DYNAMIC)
2384 rtm->rtm_protocol = RTPROT_REDIRECT;
2385 else if (rt->rt6i_flags & RTF_ADDRCONF)
2386 rtm->rtm_protocol = RTPROT_KERNEL;
2387 else if (rt->rt6i_flags & RTF_DEFAULT)
2388 rtm->rtm_protocol = RTPROT_RA;
2390 if (rt->rt6i_flags & RTF_CACHE)
2391 rtm->rtm_flags |= RTM_F_CLONED;
2394 if (nla_put(skb, RTA_DST, 16, dst))
2395 goto nla_put_failure;
2396 rtm->rtm_dst_len = 128;
2397 } else if (rtm->rtm_dst_len)
2398 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2399 goto nla_put_failure;
2400 #ifdef CONFIG_IPV6_SUBTREES
2402 if (nla_put(skb, RTA_SRC, 16, src))
2403 goto nla_put_failure;
2404 rtm->rtm_src_len = 128;
2405 } else if (rtm->rtm_src_len &&
2406 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2407 goto nla_put_failure;
2410 #ifdef CONFIG_IPV6_MROUTE
2411 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2412 int err = ip6mr_get_route(net, skb, rtm, nowait);
2417 goto nla_put_failure;
2419 if (err == -EMSGSIZE)
2420 goto nla_put_failure;
2425 if (nla_put_u32(skb, RTA_IIF, iif))
2426 goto nla_put_failure;
2428 struct in6_addr saddr_buf;
2429 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2430 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2431 goto nla_put_failure;
2434 if (rt->rt6i_prefsrc.plen) {
2435 struct in6_addr saddr_buf;
2436 saddr_buf = rt->rt6i_prefsrc.addr;
2437 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2438 goto nla_put_failure;
2441 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2442 goto nla_put_failure;
2445 n = dst_get_neighbour_noref(&rt->dst);
2447 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0) {
2449 goto nla_put_failure;
2455 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2456 goto nla_put_failure;
2457 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2458 goto nla_put_failure;
2459 if (!(rt->rt6i_flags & RTF_EXPIRES))
2461 else if (rt->dst.expires - jiffies < INT_MAX)
2462 expires = rt->dst.expires - jiffies;
2467 if (rt6_has_peer(rt))
2468 peer = rt6_peer_ptr(rt);
2470 if (peer && peer->tcp_ts_stamp) {
2472 tsage = get_seconds() - peer->tcp_ts_stamp;
2475 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, ts, tsage,
2476 expires, rt->dst.error) < 0)
2477 goto nla_put_failure;
2479 return nlmsg_end(skb, nlh);
2482 nlmsg_cancel(skb, nlh);
2486 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2488 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2491 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2492 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2493 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2497 return rt6_fill_node(arg->net,
2498 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2499 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2500 prefix, 0, NLM_F_MULTI);
2503 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2505 struct net *net = sock_net(in_skb->sk);
2506 struct nlattr *tb[RTA_MAX+1];
2507 struct rt6_info *rt;
2508 struct sk_buff *skb;
2511 int err, iif = 0, oif = 0;
2513 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2518 memset(&fl6, 0, sizeof(fl6));
2521 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2524 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2528 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2531 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2535 iif = nla_get_u32(tb[RTA_IIF]);
2538 oif = nla_get_u32(tb[RTA_OIF]);
2541 struct net_device *dev;
2544 dev = __dev_get_by_index(net, iif);
2550 fl6.flowi6_iif = iif;
2552 if (!ipv6_addr_any(&fl6.saddr))
2553 flags |= RT6_LOOKUP_F_HAS_SADDR;
2555 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2558 fl6.flowi6_oif = oif;
2560 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2563 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2565 dst_release(&rt->dst);
2570 /* Reserve room for dummy headers, this skb can pass
2571 through good chunk of routing engine.
2573 skb_reset_mac_header(skb);
2574 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2576 skb_dst_set(skb, &rt->dst);
2578 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2579 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2580 nlh->nlmsg_seq, 0, 0, 0);
2586 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2591 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2593 struct sk_buff *skb;
2594 struct net *net = info->nl_net;
2599 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2601 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2605 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2606 event, info->pid, seq, 0, 0, 0);
2608 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2609 WARN_ON(err == -EMSGSIZE);
2613 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2614 info->nlh, gfp_any());
2618 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2621 static int ip6_route_dev_notify(struct notifier_block *this,
2622 unsigned long event, void *data)
2624 struct net_device *dev = (struct net_device *)data;
2625 struct net *net = dev_net(dev);
2627 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2628 net->ipv6.ip6_null_entry->dst.dev = dev;
2629 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2630 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2631 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2632 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2633 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2634 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2645 #ifdef CONFIG_PROC_FS
2656 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2658 struct seq_file *m = p_arg;
2659 struct neighbour *n;
2661 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2663 #ifdef CONFIG_IPV6_SUBTREES
2664 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2666 seq_puts(m, "00000000000000000000000000000000 00 ");
2669 n = dst_get_neighbour_noref(&rt->dst);
2671 seq_printf(m, "%pi6", n->primary_key);
2673 seq_puts(m, "00000000000000000000000000000000");
2676 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2677 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2678 rt->dst.__use, rt->rt6i_flags,
2679 rt->dst.dev ? rt->dst.dev->name : "");
2683 static int ipv6_route_show(struct seq_file *m, void *v)
2685 struct net *net = (struct net *)m->private;
2686 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2690 static int ipv6_route_open(struct inode *inode, struct file *file)
2692 return single_open_net(inode, file, ipv6_route_show);
2695 static const struct file_operations ipv6_route_proc_fops = {
2696 .owner = THIS_MODULE,
2697 .open = ipv6_route_open,
2699 .llseek = seq_lseek,
2700 .release = single_release_net,
2703 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2705 struct net *net = (struct net *)seq->private;
2706 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2707 net->ipv6.rt6_stats->fib_nodes,
2708 net->ipv6.rt6_stats->fib_route_nodes,
2709 net->ipv6.rt6_stats->fib_rt_alloc,
2710 net->ipv6.rt6_stats->fib_rt_entries,
2711 net->ipv6.rt6_stats->fib_rt_cache,
2712 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2713 net->ipv6.rt6_stats->fib_discarded_routes);
2718 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2720 return single_open_net(inode, file, rt6_stats_seq_show);
2723 static const struct file_operations rt6_stats_seq_fops = {
2724 .owner = THIS_MODULE,
2725 .open = rt6_stats_seq_open,
2727 .llseek = seq_lseek,
2728 .release = single_release_net,
2730 #endif /* CONFIG_PROC_FS */
2732 #ifdef CONFIG_SYSCTL
2735 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2736 void __user *buffer, size_t *lenp, loff_t *ppos)
2743 net = (struct net *)ctl->extra1;
2744 delay = net->ipv6.sysctl.flush_delay;
2745 proc_dointvec(ctl, write, buffer, lenp, ppos);
2746 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2750 ctl_table ipv6_route_table_template[] = {
2752 .procname = "flush",
2753 .data = &init_net.ipv6.sysctl.flush_delay,
2754 .maxlen = sizeof(int),
2756 .proc_handler = ipv6_sysctl_rtcache_flush
2759 .procname = "gc_thresh",
2760 .data = &ip6_dst_ops_template.gc_thresh,
2761 .maxlen = sizeof(int),
2763 .proc_handler = proc_dointvec,
2766 .procname = "max_size",
2767 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2768 .maxlen = sizeof(int),
2770 .proc_handler = proc_dointvec,
2773 .procname = "gc_min_interval",
2774 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2775 .maxlen = sizeof(int),
2777 .proc_handler = proc_dointvec_jiffies,
2780 .procname = "gc_timeout",
2781 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2782 .maxlen = sizeof(int),
2784 .proc_handler = proc_dointvec_jiffies,
2787 .procname = "gc_interval",
2788 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2789 .maxlen = sizeof(int),
2791 .proc_handler = proc_dointvec_jiffies,
2794 .procname = "gc_elasticity",
2795 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2796 .maxlen = sizeof(int),
2798 .proc_handler = proc_dointvec,
2801 .procname = "mtu_expires",
2802 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2803 .maxlen = sizeof(int),
2805 .proc_handler = proc_dointvec_jiffies,
2808 .procname = "min_adv_mss",
2809 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2810 .maxlen = sizeof(int),
2812 .proc_handler = proc_dointvec,
2815 .procname = "gc_min_interval_ms",
2816 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2817 .maxlen = sizeof(int),
2819 .proc_handler = proc_dointvec_ms_jiffies,
2824 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2826 struct ctl_table *table;
2828 table = kmemdup(ipv6_route_table_template,
2829 sizeof(ipv6_route_table_template),
2833 table[0].data = &net->ipv6.sysctl.flush_delay;
2834 table[0].extra1 = net;
2835 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2836 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2837 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2838 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2839 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2840 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2841 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2842 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2843 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2850 static int __net_init ip6_route_net_init(struct net *net)
2854 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2855 sizeof(net->ipv6.ip6_dst_ops));
2857 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2858 goto out_ip6_dst_ops;
2860 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2861 sizeof(*net->ipv6.ip6_null_entry),
2863 if (!net->ipv6.ip6_null_entry)
2864 goto out_ip6_dst_entries;
2865 net->ipv6.ip6_null_entry->dst.path =
2866 (struct dst_entry *)net->ipv6.ip6_null_entry;
2867 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2868 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2869 ip6_template_metrics, true);
2871 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2872 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2873 sizeof(*net->ipv6.ip6_prohibit_entry),
2875 if (!net->ipv6.ip6_prohibit_entry)
2876 goto out_ip6_null_entry;
2877 net->ipv6.ip6_prohibit_entry->dst.path =
2878 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2879 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2880 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2881 ip6_template_metrics, true);
2883 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2884 sizeof(*net->ipv6.ip6_blk_hole_entry),
2886 if (!net->ipv6.ip6_blk_hole_entry)
2887 goto out_ip6_prohibit_entry;
2888 net->ipv6.ip6_blk_hole_entry->dst.path =
2889 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2890 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2891 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2892 ip6_template_metrics, true);
2895 net->ipv6.sysctl.flush_delay = 0;
2896 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2897 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2898 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2899 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2900 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2901 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2902 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2904 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2910 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2911 out_ip6_prohibit_entry:
2912 kfree(net->ipv6.ip6_prohibit_entry);
2914 kfree(net->ipv6.ip6_null_entry);
2916 out_ip6_dst_entries:
2917 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2922 static void __net_exit ip6_route_net_exit(struct net *net)
2924 kfree(net->ipv6.ip6_null_entry);
2925 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2926 kfree(net->ipv6.ip6_prohibit_entry);
2927 kfree(net->ipv6.ip6_blk_hole_entry);
2929 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2932 static int __net_init ip6_route_net_init_late(struct net *net)
2934 #ifdef CONFIG_PROC_FS
2935 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2936 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2941 static void __net_exit ip6_route_net_exit_late(struct net *net)
2943 #ifdef CONFIG_PROC_FS
2944 proc_net_remove(net, "ipv6_route");
2945 proc_net_remove(net, "rt6_stats");
2949 static struct pernet_operations ip6_route_net_ops = {
2950 .init = ip6_route_net_init,
2951 .exit = ip6_route_net_exit,
2954 static int __net_init ipv6_inetpeer_init(struct net *net)
2956 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
2960 inet_peer_base_init(bp);
2961 net->ipv6.peers = bp;
2965 static void __net_exit ipv6_inetpeer_exit(struct net *net)
2967 struct inet_peer_base *bp = net->ipv6.peers;
2969 net->ipv6.peers = NULL;
2970 inetpeer_invalidate_tree(bp);
2974 static struct pernet_operations ipv6_inetpeer_ops = {
2975 .init = ipv6_inetpeer_init,
2976 .exit = ipv6_inetpeer_exit,
2979 static struct pernet_operations ip6_route_net_late_ops = {
2980 .init = ip6_route_net_init_late,
2981 .exit = ip6_route_net_exit_late,
2984 static struct notifier_block ip6_route_dev_notifier = {
2985 .notifier_call = ip6_route_dev_notify,
2989 int __init ip6_route_init(void)
2994 ip6_dst_ops_template.kmem_cachep =
2995 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2996 SLAB_HWCACHE_ALIGN, NULL);
2997 if (!ip6_dst_ops_template.kmem_cachep)
3000 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3002 goto out_kmem_cache;
3004 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3006 goto out_dst_entries;
3008 ret = register_pernet_subsys(&ip6_route_net_ops);
3010 goto out_register_inetpeer;
3012 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3014 /* Registering of the loopback is done before this portion of code,
3015 * the loopback reference in rt6_info will not be taken, do it
3016 * manually for init_net */
3017 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3018 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3019 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3020 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3021 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3022 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3023 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3027 goto out_register_subsys;
3033 ret = fib6_rules_init();
3037 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3039 goto fib6_rules_init;
3042 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3043 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3044 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3045 goto out_register_late_subsys;
3047 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3049 goto out_register_late_subsys;
3054 out_register_late_subsys:
3055 unregister_pernet_subsys(&ip6_route_net_late_ops);
3057 fib6_rules_cleanup();
3062 out_register_subsys:
3063 unregister_pernet_subsys(&ip6_route_net_ops);
3064 out_register_inetpeer:
3065 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3067 dst_entries_destroy(&ip6_dst_blackhole_ops);
3069 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3073 void ip6_route_cleanup(void)
3075 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3076 unregister_pernet_subsys(&ip6_route_net_late_ops);
3077 fib6_rules_cleanup();
3080 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3081 unregister_pernet_subsys(&ip6_route_net_ops);
3082 dst_entries_destroy(&ip6_dst_blackhole_ops);
3083 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);