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);
1691 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1693 if (rt == net->ipv6.ip6_null_entry) {
1694 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1699 * We have finally decided to accept it.
1702 neigh_update(neigh, lladdr, NUD_STALE,
1703 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1704 NEIGH_UPDATE_F_OVERRIDE|
1705 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1706 NEIGH_UPDATE_F_ISROUTER))
1710 * Redirect received -> path was valid.
1711 * Look, redirects are sent only in response to data packets,
1712 * so that this nexthop apparently is reachable. --ANK
1714 dst_confirm(&rt->dst);
1716 /* Duplicate redirect: silently ignore. */
1717 if (neigh == dst_get_neighbour_noref_raw(&rt->dst))
1720 nrt = ip6_rt_copy(rt, dest);
1724 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1726 nrt->rt6i_flags &= ~RTF_GATEWAY;
1728 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
1729 dst_set_neighbour(&nrt->dst, neigh_clone(neigh));
1731 if (ip6_ins_rt(nrt))
1734 netevent.old = &rt->dst;
1735 netevent.new = &nrt->dst;
1736 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1738 if (rt->rt6i_flags & RTF_CACHE) {
1744 dst_release(&rt->dst);
1748 * Misc support functions
1751 static struct rt6_info *ip6_rt_copy(struct rt6_info *ort,
1752 const struct in6_addr *dest)
1754 struct net *net = dev_net(ort->dst.dev);
1755 struct rt6_info *rt = ip6_dst_alloc(net, ort->dst.dev, 0,
1759 rt->dst.input = ort->dst.input;
1760 rt->dst.output = ort->dst.output;
1761 rt->dst.flags |= DST_HOST;
1763 rt->rt6i_dst.addr = *dest;
1764 rt->rt6i_dst.plen = 128;
1765 dst_copy_metrics(&rt->dst, &ort->dst);
1766 rt->dst.error = ort->dst.error;
1767 rt->rt6i_idev = ort->rt6i_idev;
1769 in6_dev_hold(rt->rt6i_idev);
1770 rt->dst.lastuse = jiffies;
1772 rt->rt6i_gateway = ort->rt6i_gateway;
1773 rt->rt6i_flags = ort->rt6i_flags;
1774 if ((ort->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) ==
1775 (RTF_DEFAULT | RTF_ADDRCONF))
1776 rt6_set_from(rt, ort);
1778 rt6_clean_expires(rt);
1779 rt->rt6i_metric = 0;
1781 #ifdef CONFIG_IPV6_SUBTREES
1782 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1784 memcpy(&rt->rt6i_prefsrc, &ort->rt6i_prefsrc, sizeof(struct rt6key));
1785 rt->rt6i_table = ort->rt6i_table;
1790 #ifdef CONFIG_IPV6_ROUTE_INFO
1791 static struct rt6_info *rt6_get_route_info(struct net *net,
1792 const struct in6_addr *prefix, int prefixlen,
1793 const struct in6_addr *gwaddr, int ifindex)
1795 struct fib6_node *fn;
1796 struct rt6_info *rt = NULL;
1797 struct fib6_table *table;
1799 table = fib6_get_table(net, RT6_TABLE_INFO);
1803 write_lock_bh(&table->tb6_lock);
1804 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1808 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1809 if (rt->dst.dev->ifindex != ifindex)
1811 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1813 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1819 write_unlock_bh(&table->tb6_lock);
1823 static struct rt6_info *rt6_add_route_info(struct net *net,
1824 const struct in6_addr *prefix, int prefixlen,
1825 const struct in6_addr *gwaddr, int ifindex,
1828 struct fib6_config cfg = {
1829 .fc_table = RT6_TABLE_INFO,
1830 .fc_metric = IP6_RT_PRIO_USER,
1831 .fc_ifindex = ifindex,
1832 .fc_dst_len = prefixlen,
1833 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1834 RTF_UP | RTF_PREF(pref),
1836 .fc_nlinfo.nlh = NULL,
1837 .fc_nlinfo.nl_net = net,
1840 cfg.fc_dst = *prefix;
1841 cfg.fc_gateway = *gwaddr;
1843 /* We should treat it as a default route if prefix length is 0. */
1845 cfg.fc_flags |= RTF_DEFAULT;
1847 ip6_route_add(&cfg);
1849 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1853 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
1855 struct rt6_info *rt;
1856 struct fib6_table *table;
1858 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1862 write_lock_bh(&table->tb6_lock);
1863 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1864 if (dev == rt->dst.dev &&
1865 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1866 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1871 write_unlock_bh(&table->tb6_lock);
1875 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
1876 struct net_device *dev,
1879 struct fib6_config cfg = {
1880 .fc_table = RT6_TABLE_DFLT,
1881 .fc_metric = IP6_RT_PRIO_USER,
1882 .fc_ifindex = dev->ifindex,
1883 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1884 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1886 .fc_nlinfo.nlh = NULL,
1887 .fc_nlinfo.nl_net = dev_net(dev),
1890 cfg.fc_gateway = *gwaddr;
1892 ip6_route_add(&cfg);
1894 return rt6_get_dflt_router(gwaddr, dev);
1897 void rt6_purge_dflt_routers(struct net *net)
1899 struct rt6_info *rt;
1900 struct fib6_table *table;
1902 /* NOTE: Keep consistent with rt6_get_dflt_router */
1903 table = fib6_get_table(net, RT6_TABLE_DFLT);
1908 read_lock_bh(&table->tb6_lock);
1909 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1910 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1912 read_unlock_bh(&table->tb6_lock);
1917 read_unlock_bh(&table->tb6_lock);
1920 static void rtmsg_to_fib6_config(struct net *net,
1921 struct in6_rtmsg *rtmsg,
1922 struct fib6_config *cfg)
1924 memset(cfg, 0, sizeof(*cfg));
1926 cfg->fc_table = RT6_TABLE_MAIN;
1927 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1928 cfg->fc_metric = rtmsg->rtmsg_metric;
1929 cfg->fc_expires = rtmsg->rtmsg_info;
1930 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1931 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1932 cfg->fc_flags = rtmsg->rtmsg_flags;
1934 cfg->fc_nlinfo.nl_net = net;
1936 cfg->fc_dst = rtmsg->rtmsg_dst;
1937 cfg->fc_src = rtmsg->rtmsg_src;
1938 cfg->fc_gateway = rtmsg->rtmsg_gateway;
1941 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1943 struct fib6_config cfg;
1944 struct in6_rtmsg rtmsg;
1948 case SIOCADDRT: /* Add a route */
1949 case SIOCDELRT: /* Delete a route */
1950 if (!capable(CAP_NET_ADMIN))
1952 err = copy_from_user(&rtmsg, arg,
1953 sizeof(struct in6_rtmsg));
1957 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1962 err = ip6_route_add(&cfg);
1965 err = ip6_route_del(&cfg);
1979 * Drop the packet on the floor
1982 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1985 struct dst_entry *dst = skb_dst(skb);
1986 switch (ipstats_mib_noroutes) {
1987 case IPSTATS_MIB_INNOROUTES:
1988 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1989 if (type == IPV6_ADDR_ANY) {
1990 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1991 IPSTATS_MIB_INADDRERRORS);
1995 case IPSTATS_MIB_OUTNOROUTES:
1996 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1997 ipstats_mib_noroutes);
2000 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2005 static int ip6_pkt_discard(struct sk_buff *skb)
2007 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2010 static int ip6_pkt_discard_out(struct sk_buff *skb)
2012 skb->dev = skb_dst(skb)->dev;
2013 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2016 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2018 static int ip6_pkt_prohibit(struct sk_buff *skb)
2020 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2023 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
2025 skb->dev = skb_dst(skb)->dev;
2026 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2032 * Allocate a dst for local (unicast / anycast) address.
2035 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2036 const struct in6_addr *addr,
2039 struct net *net = dev_net(idev->dev);
2040 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev, 0, NULL);
2044 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2045 return ERR_PTR(-ENOMEM);
2050 rt->dst.flags |= DST_HOST;
2051 rt->dst.input = ip6_input;
2052 rt->dst.output = ip6_output;
2053 rt->rt6i_idev = idev;
2054 rt->dst.obsolete = -1;
2056 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2058 rt->rt6i_flags |= RTF_ANYCAST;
2060 rt->rt6i_flags |= RTF_LOCAL;
2061 err = rt6_bind_neighbour(rt, rt->dst.dev);
2064 return ERR_PTR(err);
2067 rt->rt6i_dst.addr = *addr;
2068 rt->rt6i_dst.plen = 128;
2069 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2071 atomic_set(&rt->dst.__refcnt, 1);
2076 int ip6_route_get_saddr(struct net *net,
2077 struct rt6_info *rt,
2078 const struct in6_addr *daddr,
2080 struct in6_addr *saddr)
2082 struct inet6_dev *idev = ip6_dst_idev((struct dst_entry*)rt);
2084 if (rt->rt6i_prefsrc.plen)
2085 *saddr = rt->rt6i_prefsrc.addr;
2087 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2088 daddr, prefs, saddr);
2092 /* remove deleted ip from prefsrc entries */
2093 struct arg_dev_net_ip {
2094 struct net_device *dev;
2096 struct in6_addr *addr;
2099 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2101 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2102 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2103 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2105 if (((void *)rt->dst.dev == dev || !dev) &&
2106 rt != net->ipv6.ip6_null_entry &&
2107 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2108 /* remove prefsrc entry */
2109 rt->rt6i_prefsrc.plen = 0;
2114 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2116 struct net *net = dev_net(ifp->idev->dev);
2117 struct arg_dev_net_ip adni = {
2118 .dev = ifp->idev->dev,
2122 fib6_clean_all(net, fib6_remove_prefsrc, 0, &adni);
2125 struct arg_dev_net {
2126 struct net_device *dev;
2130 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2132 const struct arg_dev_net *adn = arg;
2133 const struct net_device *dev = adn->dev;
2135 if ((rt->dst.dev == dev || !dev) &&
2136 rt != adn->net->ipv6.ip6_null_entry)
2142 void rt6_ifdown(struct net *net, struct net_device *dev)
2144 struct arg_dev_net adn = {
2149 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2150 icmp6_clean_all(fib6_ifdown, &adn);
2153 struct rt6_mtu_change_arg {
2154 struct net_device *dev;
2158 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2160 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2161 struct inet6_dev *idev;
2163 /* In IPv6 pmtu discovery is not optional,
2164 so that RTAX_MTU lock cannot disable it.
2165 We still use this lock to block changes
2166 caused by addrconf/ndisc.
2169 idev = __in6_dev_get(arg->dev);
2173 /* For administrative MTU increase, there is no way to discover
2174 IPv6 PMTU increase, so PMTU increase should be updated here.
2175 Since RFC 1981 doesn't include administrative MTU increase
2176 update PMTU increase is a MUST. (i.e. jumbo frame)
2179 If new MTU is less than route PMTU, this new MTU will be the
2180 lowest MTU in the path, update the route PMTU to reflect PMTU
2181 decreases; if new MTU is greater than route PMTU, and the
2182 old MTU is the lowest MTU in the path, update the route PMTU
2183 to reflect the increase. In this case if the other nodes' MTU
2184 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2187 if (rt->dst.dev == arg->dev &&
2188 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2189 (dst_mtu(&rt->dst) >= arg->mtu ||
2190 (dst_mtu(&rt->dst) < arg->mtu &&
2191 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2192 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2197 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2199 struct rt6_mtu_change_arg arg = {
2204 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2207 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2208 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2209 [RTA_OIF] = { .type = NLA_U32 },
2210 [RTA_IIF] = { .type = NLA_U32 },
2211 [RTA_PRIORITY] = { .type = NLA_U32 },
2212 [RTA_METRICS] = { .type = NLA_NESTED },
2215 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2216 struct fib6_config *cfg)
2219 struct nlattr *tb[RTA_MAX+1];
2222 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2227 rtm = nlmsg_data(nlh);
2228 memset(cfg, 0, sizeof(*cfg));
2230 cfg->fc_table = rtm->rtm_table;
2231 cfg->fc_dst_len = rtm->rtm_dst_len;
2232 cfg->fc_src_len = rtm->rtm_src_len;
2233 cfg->fc_flags = RTF_UP;
2234 cfg->fc_protocol = rtm->rtm_protocol;
2236 if (rtm->rtm_type == RTN_UNREACHABLE)
2237 cfg->fc_flags |= RTF_REJECT;
2239 if (rtm->rtm_type == RTN_LOCAL)
2240 cfg->fc_flags |= RTF_LOCAL;
2242 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2243 cfg->fc_nlinfo.nlh = nlh;
2244 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2246 if (tb[RTA_GATEWAY]) {
2247 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2248 cfg->fc_flags |= RTF_GATEWAY;
2252 int plen = (rtm->rtm_dst_len + 7) >> 3;
2254 if (nla_len(tb[RTA_DST]) < plen)
2257 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2261 int plen = (rtm->rtm_src_len + 7) >> 3;
2263 if (nla_len(tb[RTA_SRC]) < plen)
2266 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2269 if (tb[RTA_PREFSRC])
2270 nla_memcpy(&cfg->fc_prefsrc, tb[RTA_PREFSRC], 16);
2273 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2275 if (tb[RTA_PRIORITY])
2276 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2278 if (tb[RTA_METRICS]) {
2279 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2280 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2284 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2291 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2293 struct fib6_config cfg;
2296 err = rtm_to_fib6_config(skb, nlh, &cfg);
2300 return ip6_route_del(&cfg);
2303 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2305 struct fib6_config cfg;
2308 err = rtm_to_fib6_config(skb, nlh, &cfg);
2312 return ip6_route_add(&cfg);
2315 static inline size_t rt6_nlmsg_size(void)
2317 return NLMSG_ALIGN(sizeof(struct rtmsg))
2318 + nla_total_size(16) /* RTA_SRC */
2319 + nla_total_size(16) /* RTA_DST */
2320 + nla_total_size(16) /* RTA_GATEWAY */
2321 + nla_total_size(16) /* RTA_PREFSRC */
2322 + nla_total_size(4) /* RTA_TABLE */
2323 + nla_total_size(4) /* RTA_IIF */
2324 + nla_total_size(4) /* RTA_OIF */
2325 + nla_total_size(4) /* RTA_PRIORITY */
2326 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2327 + nla_total_size(sizeof(struct rta_cacheinfo));
2330 static int rt6_fill_node(struct net *net,
2331 struct sk_buff *skb, struct rt6_info *rt,
2332 struct in6_addr *dst, struct in6_addr *src,
2333 int iif, int type, u32 pid, u32 seq,
2334 int prefix, int nowait, unsigned int flags)
2336 const struct inet_peer *peer;
2338 struct nlmsghdr *nlh;
2341 struct neighbour *n;
2344 if (prefix) { /* user wants prefix routes only */
2345 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2346 /* success since this is not a prefix route */
2351 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2355 rtm = nlmsg_data(nlh);
2356 rtm->rtm_family = AF_INET6;
2357 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2358 rtm->rtm_src_len = rt->rt6i_src.plen;
2361 table = rt->rt6i_table->tb6_id;
2363 table = RT6_TABLE_UNSPEC;
2364 rtm->rtm_table = table;
2365 if (nla_put_u32(skb, RTA_TABLE, table))
2366 goto nla_put_failure;
2367 if (rt->rt6i_flags & RTF_REJECT)
2368 rtm->rtm_type = RTN_UNREACHABLE;
2369 else if (rt->rt6i_flags & RTF_LOCAL)
2370 rtm->rtm_type = RTN_LOCAL;
2371 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
2372 rtm->rtm_type = RTN_LOCAL;
2374 rtm->rtm_type = RTN_UNICAST;
2376 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2377 rtm->rtm_protocol = rt->rt6i_protocol;
2378 if (rt->rt6i_flags & RTF_DYNAMIC)
2379 rtm->rtm_protocol = RTPROT_REDIRECT;
2380 else if (rt->rt6i_flags & RTF_ADDRCONF)
2381 rtm->rtm_protocol = RTPROT_KERNEL;
2382 else if (rt->rt6i_flags & RTF_DEFAULT)
2383 rtm->rtm_protocol = RTPROT_RA;
2385 if (rt->rt6i_flags & RTF_CACHE)
2386 rtm->rtm_flags |= RTM_F_CLONED;
2389 if (nla_put(skb, RTA_DST, 16, dst))
2390 goto nla_put_failure;
2391 rtm->rtm_dst_len = 128;
2392 } else if (rtm->rtm_dst_len)
2393 if (nla_put(skb, RTA_DST, 16, &rt->rt6i_dst.addr))
2394 goto nla_put_failure;
2395 #ifdef CONFIG_IPV6_SUBTREES
2397 if (nla_put(skb, RTA_SRC, 16, src))
2398 goto nla_put_failure;
2399 rtm->rtm_src_len = 128;
2400 } else if (rtm->rtm_src_len &&
2401 nla_put(skb, RTA_SRC, 16, &rt->rt6i_src.addr))
2402 goto nla_put_failure;
2405 #ifdef CONFIG_IPV6_MROUTE
2406 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2407 int err = ip6mr_get_route(net, skb, rtm, nowait);
2412 goto nla_put_failure;
2414 if (err == -EMSGSIZE)
2415 goto nla_put_failure;
2420 if (nla_put_u32(skb, RTA_IIF, iif))
2421 goto nla_put_failure;
2423 struct in6_addr saddr_buf;
2424 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
2425 nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2426 goto nla_put_failure;
2429 if (rt->rt6i_prefsrc.plen) {
2430 struct in6_addr saddr_buf;
2431 saddr_buf = rt->rt6i_prefsrc.addr;
2432 if (nla_put(skb, RTA_PREFSRC, 16, &saddr_buf))
2433 goto nla_put_failure;
2436 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2437 goto nla_put_failure;
2440 n = dst_get_neighbour_noref(&rt->dst);
2442 if (nla_put(skb, RTA_GATEWAY, 16, &n->primary_key) < 0) {
2444 goto nla_put_failure;
2450 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
2451 goto nla_put_failure;
2452 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
2453 goto nla_put_failure;
2454 if (!(rt->rt6i_flags & RTF_EXPIRES))
2456 else if (rt->dst.expires - jiffies < INT_MAX)
2457 expires = rt->dst.expires - jiffies;
2462 if (rt6_has_peer(rt))
2463 peer = rt6_peer_ptr(rt);
2465 if (peer && peer->tcp_ts_stamp) {
2467 tsage = get_seconds() - peer->tcp_ts_stamp;
2470 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, ts, tsage,
2471 expires, rt->dst.error) < 0)
2472 goto nla_put_failure;
2474 return nlmsg_end(skb, nlh);
2477 nlmsg_cancel(skb, nlh);
2481 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2483 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2486 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2487 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2488 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2492 return rt6_fill_node(arg->net,
2493 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2494 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2495 prefix, 0, NLM_F_MULTI);
2498 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2500 struct net *net = sock_net(in_skb->sk);
2501 struct nlattr *tb[RTA_MAX+1];
2502 struct rt6_info *rt;
2503 struct sk_buff *skb;
2506 int err, iif = 0, oif = 0;
2508 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2513 memset(&fl6, 0, sizeof(fl6));
2516 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2519 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
2523 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2526 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
2530 iif = nla_get_u32(tb[RTA_IIF]);
2533 oif = nla_get_u32(tb[RTA_OIF]);
2536 struct net_device *dev;
2539 dev = __dev_get_by_index(net, iif);
2545 fl6.flowi6_iif = iif;
2547 if (!ipv6_addr_any(&fl6.saddr))
2548 flags |= RT6_LOOKUP_F_HAS_SADDR;
2550 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
2553 fl6.flowi6_oif = oif;
2555 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
2558 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2560 dst_release(&rt->dst);
2565 /* Reserve room for dummy headers, this skb can pass
2566 through good chunk of routing engine.
2568 skb_reset_mac_header(skb);
2569 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2571 skb_dst_set(skb, &rt->dst);
2573 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
2574 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2575 nlh->nlmsg_seq, 0, 0, 0);
2581 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2586 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2588 struct sk_buff *skb;
2589 struct net *net = info->nl_net;
2594 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2596 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2600 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2601 event, info->pid, seq, 0, 0, 0);
2603 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2604 WARN_ON(err == -EMSGSIZE);
2608 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2609 info->nlh, gfp_any());
2613 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2616 static int ip6_route_dev_notify(struct notifier_block *this,
2617 unsigned long event, void *data)
2619 struct net_device *dev = (struct net_device *)data;
2620 struct net *net = dev_net(dev);
2622 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2623 net->ipv6.ip6_null_entry->dst.dev = dev;
2624 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2625 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2626 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2627 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2628 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2629 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2640 #ifdef CONFIG_PROC_FS
2651 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2653 struct seq_file *m = p_arg;
2654 struct neighbour *n;
2656 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2658 #ifdef CONFIG_IPV6_SUBTREES
2659 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2661 seq_puts(m, "00000000000000000000000000000000 00 ");
2664 n = dst_get_neighbour_noref(&rt->dst);
2666 seq_printf(m, "%pi6", n->primary_key);
2668 seq_puts(m, "00000000000000000000000000000000");
2671 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2672 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2673 rt->dst.__use, rt->rt6i_flags,
2674 rt->dst.dev ? rt->dst.dev->name : "");
2678 static int ipv6_route_show(struct seq_file *m, void *v)
2680 struct net *net = (struct net *)m->private;
2681 fib6_clean_all_ro(net, rt6_info_route, 0, m);
2685 static int ipv6_route_open(struct inode *inode, struct file *file)
2687 return single_open_net(inode, file, ipv6_route_show);
2690 static const struct file_operations ipv6_route_proc_fops = {
2691 .owner = THIS_MODULE,
2692 .open = ipv6_route_open,
2694 .llseek = seq_lseek,
2695 .release = single_release_net,
2698 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2700 struct net *net = (struct net *)seq->private;
2701 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2702 net->ipv6.rt6_stats->fib_nodes,
2703 net->ipv6.rt6_stats->fib_route_nodes,
2704 net->ipv6.rt6_stats->fib_rt_alloc,
2705 net->ipv6.rt6_stats->fib_rt_entries,
2706 net->ipv6.rt6_stats->fib_rt_cache,
2707 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2708 net->ipv6.rt6_stats->fib_discarded_routes);
2713 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2715 return single_open_net(inode, file, rt6_stats_seq_show);
2718 static const struct file_operations rt6_stats_seq_fops = {
2719 .owner = THIS_MODULE,
2720 .open = rt6_stats_seq_open,
2722 .llseek = seq_lseek,
2723 .release = single_release_net,
2725 #endif /* CONFIG_PROC_FS */
2727 #ifdef CONFIG_SYSCTL
2730 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2731 void __user *buffer, size_t *lenp, loff_t *ppos)
2738 net = (struct net *)ctl->extra1;
2739 delay = net->ipv6.sysctl.flush_delay;
2740 proc_dointvec(ctl, write, buffer, lenp, ppos);
2741 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2745 ctl_table ipv6_route_table_template[] = {
2747 .procname = "flush",
2748 .data = &init_net.ipv6.sysctl.flush_delay,
2749 .maxlen = sizeof(int),
2751 .proc_handler = ipv6_sysctl_rtcache_flush
2754 .procname = "gc_thresh",
2755 .data = &ip6_dst_ops_template.gc_thresh,
2756 .maxlen = sizeof(int),
2758 .proc_handler = proc_dointvec,
2761 .procname = "max_size",
2762 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2763 .maxlen = sizeof(int),
2765 .proc_handler = proc_dointvec,
2768 .procname = "gc_min_interval",
2769 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2770 .maxlen = sizeof(int),
2772 .proc_handler = proc_dointvec_jiffies,
2775 .procname = "gc_timeout",
2776 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2777 .maxlen = sizeof(int),
2779 .proc_handler = proc_dointvec_jiffies,
2782 .procname = "gc_interval",
2783 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2784 .maxlen = sizeof(int),
2786 .proc_handler = proc_dointvec_jiffies,
2789 .procname = "gc_elasticity",
2790 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2791 .maxlen = sizeof(int),
2793 .proc_handler = proc_dointvec,
2796 .procname = "mtu_expires",
2797 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2798 .maxlen = sizeof(int),
2800 .proc_handler = proc_dointvec_jiffies,
2803 .procname = "min_adv_mss",
2804 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2805 .maxlen = sizeof(int),
2807 .proc_handler = proc_dointvec,
2810 .procname = "gc_min_interval_ms",
2811 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2812 .maxlen = sizeof(int),
2814 .proc_handler = proc_dointvec_ms_jiffies,
2819 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2821 struct ctl_table *table;
2823 table = kmemdup(ipv6_route_table_template,
2824 sizeof(ipv6_route_table_template),
2828 table[0].data = &net->ipv6.sysctl.flush_delay;
2829 table[0].extra1 = net;
2830 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2831 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2832 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2833 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2834 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2835 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2836 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2837 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2838 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2845 static int __net_init ip6_route_net_init(struct net *net)
2849 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2850 sizeof(net->ipv6.ip6_dst_ops));
2852 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2853 goto out_ip6_dst_ops;
2855 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2856 sizeof(*net->ipv6.ip6_null_entry),
2858 if (!net->ipv6.ip6_null_entry)
2859 goto out_ip6_dst_entries;
2860 net->ipv6.ip6_null_entry->dst.path =
2861 (struct dst_entry *)net->ipv6.ip6_null_entry;
2862 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2863 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2864 ip6_template_metrics, true);
2866 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2867 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2868 sizeof(*net->ipv6.ip6_prohibit_entry),
2870 if (!net->ipv6.ip6_prohibit_entry)
2871 goto out_ip6_null_entry;
2872 net->ipv6.ip6_prohibit_entry->dst.path =
2873 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2874 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2875 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2876 ip6_template_metrics, true);
2878 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2879 sizeof(*net->ipv6.ip6_blk_hole_entry),
2881 if (!net->ipv6.ip6_blk_hole_entry)
2882 goto out_ip6_prohibit_entry;
2883 net->ipv6.ip6_blk_hole_entry->dst.path =
2884 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2885 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2886 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2887 ip6_template_metrics, true);
2890 net->ipv6.sysctl.flush_delay = 0;
2891 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2892 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2893 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2894 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2895 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2896 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2897 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2899 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2905 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2906 out_ip6_prohibit_entry:
2907 kfree(net->ipv6.ip6_prohibit_entry);
2909 kfree(net->ipv6.ip6_null_entry);
2911 out_ip6_dst_entries:
2912 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2917 static void __net_exit ip6_route_net_exit(struct net *net)
2919 kfree(net->ipv6.ip6_null_entry);
2920 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2921 kfree(net->ipv6.ip6_prohibit_entry);
2922 kfree(net->ipv6.ip6_blk_hole_entry);
2924 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2927 static int __net_init ip6_route_net_init_late(struct net *net)
2929 #ifdef CONFIG_PROC_FS
2930 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2931 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2936 static void __net_exit ip6_route_net_exit_late(struct net *net)
2938 #ifdef CONFIG_PROC_FS
2939 proc_net_remove(net, "ipv6_route");
2940 proc_net_remove(net, "rt6_stats");
2944 static struct pernet_operations ip6_route_net_ops = {
2945 .init = ip6_route_net_init,
2946 .exit = ip6_route_net_exit,
2949 static int __net_init ipv6_inetpeer_init(struct net *net)
2951 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
2955 inet_peer_base_init(bp);
2956 net->ipv6.peers = bp;
2960 static void __net_exit ipv6_inetpeer_exit(struct net *net)
2962 struct inet_peer_base *bp = net->ipv6.peers;
2964 net->ipv6.peers = NULL;
2965 inetpeer_invalidate_tree(bp);
2969 static struct pernet_operations ipv6_inetpeer_ops = {
2970 .init = ipv6_inetpeer_init,
2971 .exit = ipv6_inetpeer_exit,
2974 static struct pernet_operations ip6_route_net_late_ops = {
2975 .init = ip6_route_net_init_late,
2976 .exit = ip6_route_net_exit_late,
2979 static struct notifier_block ip6_route_dev_notifier = {
2980 .notifier_call = ip6_route_dev_notify,
2984 int __init ip6_route_init(void)
2989 ip6_dst_ops_template.kmem_cachep =
2990 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2991 SLAB_HWCACHE_ALIGN, NULL);
2992 if (!ip6_dst_ops_template.kmem_cachep)
2995 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2997 goto out_kmem_cache;
2999 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3001 goto out_dst_entries;
3003 ret = register_pernet_subsys(&ip6_route_net_ops);
3005 goto out_register_inetpeer;
3007 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3009 /* Registering of the loopback is done before this portion of code,
3010 * the loopback reference in rt6_info will not be taken, do it
3011 * manually for init_net */
3012 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3013 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3014 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3015 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3016 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3017 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3018 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3022 goto out_register_subsys;
3028 ret = fib6_rules_init();
3032 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3034 goto fib6_rules_init;
3037 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3038 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3039 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3040 goto out_register_late_subsys;
3042 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3044 goto out_register_late_subsys;
3049 out_register_late_subsys:
3050 unregister_pernet_subsys(&ip6_route_net_late_ops);
3052 fib6_rules_cleanup();
3057 out_register_subsys:
3058 unregister_pernet_subsys(&ip6_route_net_ops);
3059 out_register_inetpeer:
3060 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3062 dst_entries_destroy(&ip6_dst_blackhole_ops);
3064 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3068 void ip6_route_cleanup(void)
3070 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3071 unregister_pernet_subsys(&ip6_route_net_late_ops);
3072 fib6_rules_cleanup();
3075 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3076 unregister_pernet_subsys(&ip6_route_net_ops);
3077 dst_entries_destroy(&ip6_dst_blackhole_ops);
3078 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);