2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/types.h>
30 #include <linux/times.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/route.h>
35 #include <linux/netdevice.h>
36 #include <linux/in6.h>
37 #include <linux/mroute6.h>
38 #include <linux/init.h>
39 #include <linux/if_arp.h>
40 #include <linux/proc_fs.h>
41 #include <linux/seq_file.h>
42 #include <linux/nsproxy.h>
43 #include <linux/slab.h>
44 #include <net/net_namespace.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
58 #include <asm/uaccess.h>
61 #include <linux/sysctl.h>
64 /* Set to 3 to get tracing. */
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
72 #define RT6_TRACE(x...) do { ; } while (0)
75 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
76 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
77 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
78 static unsigned int ip6_default_mtu(const struct dst_entry *dst);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static void ip6_link_failure(struct sk_buff *skb);
88 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
90 #ifdef CONFIG_IPV6_ROUTE_INFO
91 static struct rt6_info *rt6_add_route_info(struct net *net,
92 struct in6_addr *prefix, int prefixlen,
93 struct in6_addr *gwaddr, int ifindex,
95 static struct rt6_info *rt6_get_route_info(struct net *net,
96 struct in6_addr *prefix, int prefixlen,
97 struct in6_addr *gwaddr, int ifindex);
100 static struct dst_ops ip6_dst_ops_template = {
102 .protocol = cpu_to_be16(ETH_P_IPV6),
105 .check = ip6_dst_check,
106 .default_advmss = ip6_default_advmss,
107 .default_mtu = ip6_default_mtu,
108 .cow_metrics = dst_cow_metrics_generic,
109 .destroy = ip6_dst_destroy,
110 .ifdown = ip6_dst_ifdown,
111 .negative_advice = ip6_negative_advice,
112 .link_failure = ip6_link_failure,
113 .update_pmtu = ip6_rt_update_pmtu,
114 .local_out = __ip6_local_out,
117 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
121 static struct dst_ops ip6_dst_blackhole_ops = {
123 .protocol = cpu_to_be16(ETH_P_IPV6),
124 .destroy = ip6_dst_destroy,
125 .check = ip6_dst_check,
126 .update_pmtu = ip6_rt_blackhole_update_pmtu,
129 static const u32 ip6_template_metrics[RTAX_MAX] = {
130 [RTAX_HOPLIMIT - 1] = 255,
133 static struct rt6_info ip6_null_entry_template = {
135 .__refcnt = ATOMIC_INIT(1),
138 .error = -ENETUNREACH,
139 .input = ip6_pkt_discard,
140 .output = ip6_pkt_discard_out,
142 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
143 .rt6i_protocol = RTPROT_KERNEL,
144 .rt6i_metric = ~(u32) 0,
145 .rt6i_ref = ATOMIC_INIT(1),
148 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
150 static int ip6_pkt_prohibit(struct sk_buff *skb);
151 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
153 static struct rt6_info ip6_prohibit_entry_template = {
155 .__refcnt = ATOMIC_INIT(1),
159 .input = ip6_pkt_prohibit,
160 .output = ip6_pkt_prohibit_out,
162 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
163 .rt6i_protocol = RTPROT_KERNEL,
164 .rt6i_metric = ~(u32) 0,
165 .rt6i_ref = ATOMIC_INIT(1),
168 static struct rt6_info ip6_blk_hole_entry_template = {
170 .__refcnt = ATOMIC_INIT(1),
174 .input = dst_discard,
175 .output = dst_discard,
177 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
178 .rt6i_protocol = RTPROT_KERNEL,
179 .rt6i_metric = ~(u32) 0,
180 .rt6i_ref = ATOMIC_INIT(1),
185 /* allocate dst with ip6_dst_ops */
186 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
188 return (struct rt6_info *)dst_alloc(ops);
191 static void ip6_dst_destroy(struct dst_entry *dst)
193 struct rt6_info *rt = (struct rt6_info *)dst;
194 struct inet6_dev *idev = rt->rt6i_idev;
195 struct inet_peer *peer = rt->rt6i_peer;
198 rt->rt6i_idev = NULL;
201 dst_destroy_metrics_generic(dst);
203 BUG_ON(!(rt->rt6i_flags & RTF_CACHE));
204 rt->rt6i_peer = NULL;
209 void rt6_bind_peer(struct rt6_info *rt, int create)
211 struct inet_peer *peer;
213 if (WARN_ON(!(rt->rt6i_flags & RTF_CACHE)))
216 peer = inet_getpeer_v6(&rt->rt6i_dst.addr, create);
217 if (peer && cmpxchg(&rt->rt6i_peer, NULL, peer) != NULL)
221 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
224 struct rt6_info *rt = (struct rt6_info *)dst;
225 struct inet6_dev *idev = rt->rt6i_idev;
226 struct net_device *loopback_dev =
227 dev_net(dev)->loopback_dev;
229 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
230 struct inet6_dev *loopback_idev =
231 in6_dev_get(loopback_dev);
232 if (loopback_idev != NULL) {
233 rt->rt6i_idev = loopback_idev;
239 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
241 return (rt->rt6i_flags & RTF_EXPIRES) &&
242 time_after(jiffies, rt->rt6i_expires);
245 static inline int rt6_need_strict(struct in6_addr *daddr)
247 return ipv6_addr_type(daddr) &
248 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
252 * Route lookup. Any table->tb6_lock is implied.
255 static inline struct rt6_info *rt6_device_match(struct net *net,
257 struct in6_addr *saddr,
261 struct rt6_info *local = NULL;
262 struct rt6_info *sprt;
264 if (!oif && ipv6_addr_any(saddr))
267 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
268 struct net_device *dev = sprt->rt6i_dev;
271 if (dev->ifindex == oif)
273 if (dev->flags & IFF_LOOPBACK) {
274 if (sprt->rt6i_idev == NULL ||
275 sprt->rt6i_idev->dev->ifindex != oif) {
276 if (flags & RT6_LOOKUP_F_IFACE && oif)
278 if (local && (!oif ||
279 local->rt6i_idev->dev->ifindex == oif))
285 if (ipv6_chk_addr(net, saddr, dev,
286 flags & RT6_LOOKUP_F_IFACE))
295 if (flags & RT6_LOOKUP_F_IFACE)
296 return net->ipv6.ip6_null_entry;
302 #ifdef CONFIG_IPV6_ROUTER_PREF
303 static void rt6_probe(struct rt6_info *rt)
305 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
307 * Okay, this does not seem to be appropriate
308 * for now, however, we need to check if it
309 * is really so; aka Router Reachability Probing.
311 * Router Reachability Probe MUST be rate-limited
312 * to no more than one per minute.
314 if (!neigh || (neigh->nud_state & NUD_VALID))
316 read_lock_bh(&neigh->lock);
317 if (!(neigh->nud_state & NUD_VALID) &&
318 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
319 struct in6_addr mcaddr;
320 struct in6_addr *target;
322 neigh->updated = jiffies;
323 read_unlock_bh(&neigh->lock);
325 target = (struct in6_addr *)&neigh->primary_key;
326 addrconf_addr_solict_mult(target, &mcaddr);
327 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
329 read_unlock_bh(&neigh->lock);
332 static inline void rt6_probe(struct rt6_info *rt)
338 * Default Router Selection (RFC 2461 6.3.6)
340 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
342 struct net_device *dev = rt->rt6i_dev;
343 if (!oif || dev->ifindex == oif)
345 if ((dev->flags & IFF_LOOPBACK) &&
346 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
351 static inline int rt6_check_neigh(struct rt6_info *rt)
353 struct neighbour *neigh = rt->rt6i_nexthop;
355 if (rt->rt6i_flags & RTF_NONEXTHOP ||
356 !(rt->rt6i_flags & RTF_GATEWAY))
359 read_lock_bh(&neigh->lock);
360 if (neigh->nud_state & NUD_VALID)
362 #ifdef CONFIG_IPV6_ROUTER_PREF
363 else if (neigh->nud_state & NUD_FAILED)
368 read_unlock_bh(&neigh->lock);
374 static int rt6_score_route(struct rt6_info *rt, int oif,
379 m = rt6_check_dev(rt, oif);
380 if (!m && (strict & RT6_LOOKUP_F_IFACE))
382 #ifdef CONFIG_IPV6_ROUTER_PREF
383 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
385 n = rt6_check_neigh(rt);
386 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
391 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
392 int *mpri, struct rt6_info *match)
396 if (rt6_check_expired(rt))
399 m = rt6_score_route(rt, oif, strict);
404 if (strict & RT6_LOOKUP_F_REACHABLE)
408 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
416 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
417 struct rt6_info *rr_head,
418 u32 metric, int oif, int strict)
420 struct rt6_info *rt, *match;
424 for (rt = rr_head; rt && rt->rt6i_metric == metric;
425 rt = rt->dst.rt6_next)
426 match = find_match(rt, oif, strict, &mpri, match);
427 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
428 rt = rt->dst.rt6_next)
429 match = find_match(rt, oif, strict, &mpri, match);
434 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
436 struct rt6_info *match, *rt0;
439 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
440 __func__, fn->leaf, oif);
444 fn->rr_ptr = rt0 = fn->leaf;
446 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
449 (strict & RT6_LOOKUP_F_REACHABLE)) {
450 struct rt6_info *next = rt0->dst.rt6_next;
452 /* no entries matched; do round-robin */
453 if (!next || next->rt6i_metric != rt0->rt6i_metric)
460 RT6_TRACE("%s() => %p\n",
463 net = dev_net(rt0->rt6i_dev);
464 return match ? match : net->ipv6.ip6_null_entry;
467 #ifdef CONFIG_IPV6_ROUTE_INFO
468 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
469 struct in6_addr *gwaddr)
471 struct net *net = dev_net(dev);
472 struct route_info *rinfo = (struct route_info *) opt;
473 struct in6_addr prefix_buf, *prefix;
475 unsigned long lifetime;
478 if (len < sizeof(struct route_info)) {
482 /* Sanity check for prefix_len and length */
483 if (rinfo->length > 3) {
485 } else if (rinfo->prefix_len > 128) {
487 } else if (rinfo->prefix_len > 64) {
488 if (rinfo->length < 2) {
491 } else if (rinfo->prefix_len > 0) {
492 if (rinfo->length < 1) {
497 pref = rinfo->route_pref;
498 if (pref == ICMPV6_ROUTER_PREF_INVALID)
501 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
503 if (rinfo->length == 3)
504 prefix = (struct in6_addr *)rinfo->prefix;
506 /* this function is safe */
507 ipv6_addr_prefix(&prefix_buf,
508 (struct in6_addr *)rinfo->prefix,
510 prefix = &prefix_buf;
513 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
516 if (rt && !lifetime) {
522 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
525 rt->rt6i_flags = RTF_ROUTEINFO |
526 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
529 if (!addrconf_finite_timeout(lifetime)) {
530 rt->rt6i_flags &= ~RTF_EXPIRES;
532 rt->rt6i_expires = jiffies + HZ * lifetime;
533 rt->rt6i_flags |= RTF_EXPIRES;
535 dst_release(&rt->dst);
541 #define BACKTRACK(__net, saddr) \
543 if (rt == __net->ipv6.ip6_null_entry) { \
544 struct fib6_node *pn; \
546 if (fn->fn_flags & RTN_TL_ROOT) \
549 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
550 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
553 if (fn->fn_flags & RTN_RTINFO) \
559 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
560 struct fib6_table *table,
561 struct flowi *fl, int flags)
563 struct fib6_node *fn;
566 read_lock_bh(&table->tb6_lock);
567 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
570 rt = rt6_device_match(net, rt, &fl->fl6_src, fl->oif, flags);
571 BACKTRACK(net, &fl->fl6_src);
573 dst_use(&rt->dst, jiffies);
574 read_unlock_bh(&table->tb6_lock);
579 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
580 const struct in6_addr *saddr, int oif, int strict)
586 struct dst_entry *dst;
587 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
590 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
591 flags |= RT6_LOOKUP_F_HAS_SADDR;
594 dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
596 return (struct rt6_info *) dst;
603 EXPORT_SYMBOL(rt6_lookup);
605 /* ip6_ins_rt is called with FREE table->tb6_lock.
606 It takes new route entry, the addition fails by any reason the
607 route is freed. In any case, if caller does not hold it, it may
611 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
614 struct fib6_table *table;
616 table = rt->rt6i_table;
617 write_lock_bh(&table->tb6_lock);
618 err = fib6_add(&table->tb6_root, rt, info);
619 write_unlock_bh(&table->tb6_lock);
624 int ip6_ins_rt(struct rt6_info *rt)
626 struct nl_info info = {
627 .nl_net = dev_net(rt->rt6i_dev),
629 return __ip6_ins_rt(rt, &info);
632 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
633 struct in6_addr *saddr)
641 rt = ip6_rt_copy(ort);
644 struct neighbour *neigh;
645 int attempts = !in_softirq();
647 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
648 if (rt->rt6i_dst.plen != 128 &&
649 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
650 rt->rt6i_flags |= RTF_ANYCAST;
651 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
654 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
655 rt->rt6i_dst.plen = 128;
656 rt->rt6i_flags |= RTF_CACHE;
657 rt->dst.flags |= DST_HOST;
659 #ifdef CONFIG_IPV6_SUBTREES
660 if (rt->rt6i_src.plen && saddr) {
661 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
662 rt->rt6i_src.plen = 128;
667 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
669 struct net *net = dev_net(rt->rt6i_dev);
670 int saved_rt_min_interval =
671 net->ipv6.sysctl.ip6_rt_gc_min_interval;
672 int saved_rt_elasticity =
673 net->ipv6.sysctl.ip6_rt_gc_elasticity;
675 if (attempts-- > 0) {
676 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
677 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
679 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
681 net->ipv6.sysctl.ip6_rt_gc_elasticity =
683 net->ipv6.sysctl.ip6_rt_gc_min_interval =
684 saved_rt_min_interval;
690 "ipv6: Neighbour table overflow.\n");
694 rt->rt6i_nexthop = neigh;
701 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
703 struct rt6_info *rt = ip6_rt_copy(ort);
705 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
706 rt->rt6i_dst.plen = 128;
707 rt->rt6i_flags |= RTF_CACHE;
708 rt->dst.flags |= DST_HOST;
709 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
714 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
715 struct flowi *fl, int flags)
717 struct fib6_node *fn;
718 struct rt6_info *rt, *nrt;
722 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
724 strict |= flags & RT6_LOOKUP_F_IFACE;
727 read_lock_bh(&table->tb6_lock);
730 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
733 rt = rt6_select(fn, oif, strict | reachable);
735 BACKTRACK(net, &fl->fl6_src);
736 if (rt == net->ipv6.ip6_null_entry ||
737 rt->rt6i_flags & RTF_CACHE)
741 read_unlock_bh(&table->tb6_lock);
743 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
744 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
746 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
748 dst_release(&rt->dst);
749 rt = nrt ? : net->ipv6.ip6_null_entry;
753 err = ip6_ins_rt(nrt);
762 * Race condition! In the gap, when table->tb6_lock was
763 * released someone could insert this route. Relookup.
765 dst_release(&rt->dst);
774 read_unlock_bh(&table->tb6_lock);
776 rt->dst.lastuse = jiffies;
782 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
783 struct flowi *fl, int flags)
785 return ip6_pol_route(net, table, fl->iif, fl, flags);
788 void ip6_route_input(struct sk_buff *skb)
790 struct ipv6hdr *iph = ipv6_hdr(skb);
791 struct net *net = dev_net(skb->dev);
792 int flags = RT6_LOOKUP_F_HAS_SADDR;
794 .iif = skb->dev->ifindex,
795 .fl6_dst = iph->daddr,
796 .fl6_src = iph->saddr,
797 .fl6_flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
799 .proto = iph->nexthdr,
802 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
803 flags |= RT6_LOOKUP_F_IFACE;
805 skb_dst_set(skb, fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input));
808 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
809 struct flowi *fl, int flags)
811 return ip6_pol_route(net, table, fl->oif, fl, flags);
814 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
819 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl->fl6_dst))
820 flags |= RT6_LOOKUP_F_IFACE;
822 if (!ipv6_addr_any(&fl->fl6_src))
823 flags |= RT6_LOOKUP_F_HAS_SADDR;
825 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
827 return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
830 EXPORT_SYMBOL(ip6_route_output);
832 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
834 struct rt6_info *ort = (struct rt6_info *) *dstp;
835 struct rt6_info *rt = (struct rt6_info *)
836 dst_alloc(&ip6_dst_blackhole_ops);
837 struct dst_entry *new = NULL;
842 atomic_set(&new->__refcnt, 1);
844 new->input = dst_discard;
845 new->output = dst_discard;
847 dst_copy_metrics(new, &ort->dst);
848 new->dev = ort->dst.dev;
851 rt->rt6i_idev = ort->rt6i_idev;
853 in6_dev_hold(rt->rt6i_idev);
854 rt->rt6i_expires = 0;
856 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
857 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
860 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
861 #ifdef CONFIG_IPV6_SUBTREES
862 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
870 return new ? 0 : -ENOMEM;
872 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
875 * Destination cache support functions
878 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
882 rt = (struct rt6_info *) dst;
884 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
890 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
892 struct rt6_info *rt = (struct rt6_info *) dst;
895 if (rt->rt6i_flags & RTF_CACHE) {
896 if (rt6_check_expired(rt)) {
908 static void ip6_link_failure(struct sk_buff *skb)
912 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
914 rt = (struct rt6_info *) skb_dst(skb);
916 if (rt->rt6i_flags&RTF_CACHE) {
917 dst_set_expires(&rt->dst, 0);
918 rt->rt6i_flags |= RTF_EXPIRES;
919 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
920 rt->rt6i_node->fn_sernum = -1;
924 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
926 struct rt6_info *rt6 = (struct rt6_info*)dst;
928 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
929 rt6->rt6i_flags |= RTF_MODIFIED;
930 if (mtu < IPV6_MIN_MTU) {
931 u32 features = dst_metric(dst, RTAX_FEATURES);
933 features |= RTAX_FEATURE_ALLFRAG;
934 dst_metric_set(dst, RTAX_FEATURES, features);
936 dst_metric_set(dst, RTAX_MTU, mtu);
937 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
941 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
943 struct net_device *dev = dst->dev;
944 unsigned int mtu = dst_mtu(dst);
945 struct net *net = dev_net(dev);
947 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
949 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
950 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
953 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
954 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
955 * IPV6_MAXPLEN is also valid and means: "any MSS,
956 * rely only on pmtu discovery"
958 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
963 static unsigned int ip6_default_mtu(const struct dst_entry *dst)
965 unsigned int mtu = IPV6_MIN_MTU;
966 struct inet6_dev *idev;
969 idev = __in6_dev_get(dst->dev);
971 mtu = idev->cnf.mtu6;
977 static struct dst_entry *icmp6_dst_gc_list;
978 static DEFINE_SPINLOCK(icmp6_dst_lock);
980 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
981 struct neighbour *neigh,
982 const struct in6_addr *addr)
985 struct inet6_dev *idev = in6_dev_get(dev);
986 struct net *net = dev_net(dev);
988 if (unlikely(idev == NULL))
991 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
992 if (unlikely(rt == NULL)) {
1001 neigh = ndisc_get_neigh(dev, addr);
1007 rt->rt6i_idev = idev;
1008 rt->rt6i_nexthop = neigh;
1009 atomic_set(&rt->dst.__refcnt, 1);
1010 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 255);
1011 rt->dst.output = ip6_output;
1013 #if 0 /* there's no chance to use these for ndisc */
1014 rt->dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
1017 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1018 rt->rt6i_dst.plen = 128;
1021 spin_lock_bh(&icmp6_dst_lock);
1022 rt->dst.next = icmp6_dst_gc_list;
1023 icmp6_dst_gc_list = &rt->dst;
1024 spin_unlock_bh(&icmp6_dst_lock);
1026 fib6_force_start_gc(net);
1032 int icmp6_dst_gc(void)
1034 struct dst_entry *dst, *next, **pprev;
1039 spin_lock_bh(&icmp6_dst_lock);
1040 pprev = &icmp6_dst_gc_list;
1042 while ((dst = *pprev) != NULL) {
1043 if (!atomic_read(&dst->__refcnt)) {
1052 spin_unlock_bh(&icmp6_dst_lock);
1057 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1060 struct dst_entry *dst, **pprev;
1062 spin_lock_bh(&icmp6_dst_lock);
1063 pprev = &icmp6_dst_gc_list;
1064 while ((dst = *pprev) != NULL) {
1065 struct rt6_info *rt = (struct rt6_info *) dst;
1066 if (func(rt, arg)) {
1073 spin_unlock_bh(&icmp6_dst_lock);
1076 static int ip6_dst_gc(struct dst_ops *ops)
1078 unsigned long now = jiffies;
1079 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1080 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1081 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1082 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1083 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1084 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1087 entries = dst_entries_get_fast(ops);
1088 if (time_after(rt_last_gc + rt_min_interval, now) &&
1089 entries <= rt_max_size)
1092 net->ipv6.ip6_rt_gc_expire++;
1093 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1094 net->ipv6.ip6_rt_last_gc = now;
1095 entries = dst_entries_get_slow(ops);
1096 if (entries < ops->gc_thresh)
1097 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1099 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1100 return entries > rt_max_size;
1103 /* Clean host part of a prefix. Not necessary in radix tree,
1104 but results in cleaner routing tables.
1106 Remove it only when all the things will work!
1109 int ip6_dst_hoplimit(struct dst_entry *dst)
1111 int hoplimit = dst_metric_raw(dst, RTAX_HOPLIMIT);
1112 if (hoplimit == 0) {
1113 struct net_device *dev = dst->dev;
1114 struct inet6_dev *idev;
1117 idev = __in6_dev_get(dev);
1119 hoplimit = idev->cnf.hop_limit;
1121 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1126 EXPORT_SYMBOL(ip6_dst_hoplimit);
1132 int ip6_route_add(struct fib6_config *cfg)
1135 struct net *net = cfg->fc_nlinfo.nl_net;
1136 struct rt6_info *rt = NULL;
1137 struct net_device *dev = NULL;
1138 struct inet6_dev *idev = NULL;
1139 struct fib6_table *table;
1142 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1144 #ifndef CONFIG_IPV6_SUBTREES
1145 if (cfg->fc_src_len)
1148 if (cfg->fc_ifindex) {
1150 dev = dev_get_by_index(net, cfg->fc_ifindex);
1153 idev = in6_dev_get(dev);
1158 if (cfg->fc_metric == 0)
1159 cfg->fc_metric = IP6_RT_PRIO_USER;
1161 table = fib6_new_table(net, cfg->fc_table);
1162 if (table == NULL) {
1167 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1174 rt->dst.obsolete = -1;
1175 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1176 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1179 if (cfg->fc_protocol == RTPROT_UNSPEC)
1180 cfg->fc_protocol = RTPROT_BOOT;
1181 rt->rt6i_protocol = cfg->fc_protocol;
1183 addr_type = ipv6_addr_type(&cfg->fc_dst);
1185 if (addr_type & IPV6_ADDR_MULTICAST)
1186 rt->dst.input = ip6_mc_input;
1187 else if (cfg->fc_flags & RTF_LOCAL)
1188 rt->dst.input = ip6_input;
1190 rt->dst.input = ip6_forward;
1192 rt->dst.output = ip6_output;
1194 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1195 rt->rt6i_dst.plen = cfg->fc_dst_len;
1196 if (rt->rt6i_dst.plen == 128)
1197 rt->dst.flags = DST_HOST;
1199 #ifdef CONFIG_IPV6_SUBTREES
1200 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1201 rt->rt6i_src.plen = cfg->fc_src_len;
1204 rt->rt6i_metric = cfg->fc_metric;
1206 /* We cannot add true routes via loopback here,
1207 they would result in kernel looping; promote them to reject routes
1209 if ((cfg->fc_flags & RTF_REJECT) ||
1210 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK)
1211 && !(cfg->fc_flags&RTF_LOCAL))) {
1212 /* hold loopback dev/idev if we haven't done so. */
1213 if (dev != net->loopback_dev) {
1218 dev = net->loopback_dev;
1220 idev = in6_dev_get(dev);
1226 rt->dst.output = ip6_pkt_discard_out;
1227 rt->dst.input = ip6_pkt_discard;
1228 rt->dst.error = -ENETUNREACH;
1229 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1233 if (cfg->fc_flags & RTF_GATEWAY) {
1234 struct in6_addr *gw_addr;
1237 gw_addr = &cfg->fc_gateway;
1238 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1239 gwa_type = ipv6_addr_type(gw_addr);
1241 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1242 struct rt6_info *grt;
1244 /* IPv6 strictly inhibits using not link-local
1245 addresses as nexthop address.
1246 Otherwise, router will not able to send redirects.
1247 It is very good, but in some (rare!) circumstances
1248 (SIT, PtP, NBMA NOARP links) it is handy to allow
1249 some exceptions. --ANK
1252 if (!(gwa_type&IPV6_ADDR_UNICAST))
1255 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1257 err = -EHOSTUNREACH;
1261 if (dev != grt->rt6i_dev) {
1262 dst_release(&grt->dst);
1266 dev = grt->rt6i_dev;
1267 idev = grt->rt6i_idev;
1269 in6_dev_hold(grt->rt6i_idev);
1271 if (!(grt->rt6i_flags&RTF_GATEWAY))
1273 dst_release(&grt->dst);
1279 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1287 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1288 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1289 if (IS_ERR(rt->rt6i_nexthop)) {
1290 err = PTR_ERR(rt->rt6i_nexthop);
1291 rt->rt6i_nexthop = NULL;
1296 rt->rt6i_flags = cfg->fc_flags;
1303 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1304 int type = nla_type(nla);
1307 if (type > RTAX_MAX) {
1312 dst_metric_set(&rt->dst, type, nla_get_u32(nla));
1318 rt->rt6i_idev = idev;
1319 rt->rt6i_table = table;
1321 cfg->fc_nlinfo.nl_net = dev_net(dev);
1323 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1335 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1338 struct fib6_table *table;
1339 struct net *net = dev_net(rt->rt6i_dev);
1341 if (rt == net->ipv6.ip6_null_entry)
1344 table = rt->rt6i_table;
1345 write_lock_bh(&table->tb6_lock);
1347 err = fib6_del(rt, info);
1348 dst_release(&rt->dst);
1350 write_unlock_bh(&table->tb6_lock);
1355 int ip6_del_rt(struct rt6_info *rt)
1357 struct nl_info info = {
1358 .nl_net = dev_net(rt->rt6i_dev),
1360 return __ip6_del_rt(rt, &info);
1363 static int ip6_route_del(struct fib6_config *cfg)
1365 struct fib6_table *table;
1366 struct fib6_node *fn;
1367 struct rt6_info *rt;
1370 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1374 read_lock_bh(&table->tb6_lock);
1376 fn = fib6_locate(&table->tb6_root,
1377 &cfg->fc_dst, cfg->fc_dst_len,
1378 &cfg->fc_src, cfg->fc_src_len);
1381 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1382 if (cfg->fc_ifindex &&
1383 (rt->rt6i_dev == NULL ||
1384 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1386 if (cfg->fc_flags & RTF_GATEWAY &&
1387 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1389 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1392 read_unlock_bh(&table->tb6_lock);
1394 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1397 read_unlock_bh(&table->tb6_lock);
1405 struct ip6rd_flowi {
1407 struct in6_addr gateway;
1410 static struct rt6_info *__ip6_route_redirect(struct net *net,
1411 struct fib6_table *table,
1415 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1416 struct rt6_info *rt;
1417 struct fib6_node *fn;
1420 * Get the "current" route for this destination and
1421 * check if the redirect has come from approriate router.
1423 * RFC 2461 specifies that redirects should only be
1424 * accepted if they come from the nexthop to the target.
1425 * Due to the way the routes are chosen, this notion
1426 * is a bit fuzzy and one might need to check all possible
1430 read_lock_bh(&table->tb6_lock);
1431 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1433 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1435 * Current route is on-link; redirect is always invalid.
1437 * Seems, previous statement is not true. It could
1438 * be node, which looks for us as on-link (f.e. proxy ndisc)
1439 * But then router serving it might decide, that we should
1440 * know truth 8)8) --ANK (980726).
1442 if (rt6_check_expired(rt))
1444 if (!(rt->rt6i_flags & RTF_GATEWAY))
1446 if (fl->oif != rt->rt6i_dev->ifindex)
1448 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1454 rt = net->ipv6.ip6_null_entry;
1455 BACKTRACK(net, &fl->fl6_src);
1459 read_unlock_bh(&table->tb6_lock);
1464 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1465 struct in6_addr *src,
1466 struct in6_addr *gateway,
1467 struct net_device *dev)
1469 int flags = RT6_LOOKUP_F_HAS_SADDR;
1470 struct net *net = dev_net(dev);
1471 struct ip6rd_flowi rdfl = {
1473 .oif = dev->ifindex,
1479 ipv6_addr_copy(&rdfl.gateway, gateway);
1481 if (rt6_need_strict(dest))
1482 flags |= RT6_LOOKUP_F_IFACE;
1484 return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1485 flags, __ip6_route_redirect);
1488 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1489 struct in6_addr *saddr,
1490 struct neighbour *neigh, u8 *lladdr, int on_link)
1492 struct rt6_info *rt, *nrt = NULL;
1493 struct netevent_redirect netevent;
1494 struct net *net = dev_net(neigh->dev);
1496 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1498 if (rt == net->ipv6.ip6_null_entry) {
1499 if (net_ratelimit())
1500 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1501 "for redirect target\n");
1506 * We have finally decided to accept it.
1509 neigh_update(neigh, lladdr, NUD_STALE,
1510 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1511 NEIGH_UPDATE_F_OVERRIDE|
1512 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1513 NEIGH_UPDATE_F_ISROUTER))
1517 * Redirect received -> path was valid.
1518 * Look, redirects are sent only in response to data packets,
1519 * so that this nexthop apparently is reachable. --ANK
1521 dst_confirm(&rt->dst);
1523 /* Duplicate redirect: silently ignore. */
1524 if (neigh == rt->dst.neighbour)
1527 nrt = ip6_rt_copy(rt);
1531 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1533 nrt->rt6i_flags &= ~RTF_GATEWAY;
1535 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1536 nrt->rt6i_dst.plen = 128;
1537 nrt->dst.flags |= DST_HOST;
1539 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1540 nrt->rt6i_nexthop = neigh_clone(neigh);
1542 if (ip6_ins_rt(nrt))
1545 netevent.old = &rt->dst;
1546 netevent.new = &nrt->dst;
1547 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1549 if (rt->rt6i_flags&RTF_CACHE) {
1555 dst_release(&rt->dst);
1559 * Handle ICMP "packet too big" messages
1560 * i.e. Path MTU discovery
1563 static void rt6_do_pmtu_disc(struct in6_addr *daddr, struct in6_addr *saddr,
1564 struct net *net, u32 pmtu, int ifindex)
1566 struct rt6_info *rt, *nrt;
1569 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1573 if (rt6_check_expired(rt)) {
1578 if (pmtu >= dst_mtu(&rt->dst))
1581 if (pmtu < IPV6_MIN_MTU) {
1583 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1584 * MTU (1280) and a fragment header should always be included
1585 * after a node receiving Too Big message reporting PMTU is
1586 * less than the IPv6 Minimum Link MTU.
1588 pmtu = IPV6_MIN_MTU;
1592 /* New mtu received -> path was valid.
1593 They are sent only in response to data packets,
1594 so that this nexthop apparently is reachable. --ANK
1596 dst_confirm(&rt->dst);
1598 /* Host route. If it is static, it would be better
1599 not to override it, but add new one, so that
1600 when cache entry will expire old pmtu
1601 would return automatically.
1603 if (rt->rt6i_flags & RTF_CACHE) {
1604 dst_metric_set(&rt->dst, RTAX_MTU, pmtu);
1606 u32 features = dst_metric(&rt->dst, RTAX_FEATURES);
1607 features |= RTAX_FEATURE_ALLFRAG;
1608 dst_metric_set(&rt->dst, RTAX_FEATURES, features);
1610 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1611 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1616 Two cases are possible:
1617 1. It is connected route. Action: COW
1618 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1620 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1621 nrt = rt6_alloc_cow(rt, daddr, saddr);
1623 nrt = rt6_alloc_clone(rt, daddr);
1626 dst_metric_set(&nrt->dst, RTAX_MTU, pmtu);
1628 u32 features = dst_metric(&nrt->dst, RTAX_FEATURES);
1629 features |= RTAX_FEATURE_ALLFRAG;
1630 dst_metric_set(&nrt->dst, RTAX_FEATURES, features);
1633 /* According to RFC 1981, detecting PMTU increase shouldn't be
1634 * happened within 5 mins, the recommended timer is 10 mins.
1635 * Here this route expiration time is set to ip6_rt_mtu_expires
1636 * which is 10 mins. After 10 mins the decreased pmtu is expired
1637 * and detecting PMTU increase will be automatically happened.
1639 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1640 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1645 dst_release(&rt->dst);
1648 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1649 struct net_device *dev, u32 pmtu)
1651 struct net *net = dev_net(dev);
1654 * RFC 1981 states that a node "MUST reduce the size of the packets it
1655 * is sending along the path" that caused the Packet Too Big message.
1656 * Since it's not possible in the general case to determine which
1657 * interface was used to send the original packet, we update the MTU
1658 * on the interface that will be used to send future packets. We also
1659 * update the MTU on the interface that received the Packet Too Big in
1660 * case the original packet was forced out that interface with
1661 * SO_BINDTODEVICE or similar. This is the next best thing to the
1662 * correct behaviour, which would be to update the MTU on all
1665 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1666 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1670 * Misc support functions
1673 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1675 struct net *net = dev_net(ort->rt6i_dev);
1676 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1679 rt->dst.input = ort->dst.input;
1680 rt->dst.output = ort->dst.output;
1682 dst_copy_metrics(&rt->dst, &ort->dst);
1683 rt->dst.error = ort->dst.error;
1684 rt->dst.dev = ort->dst.dev;
1686 dev_hold(rt->dst.dev);
1687 rt->rt6i_idev = ort->rt6i_idev;
1689 in6_dev_hold(rt->rt6i_idev);
1690 rt->dst.lastuse = jiffies;
1691 rt->rt6i_expires = 0;
1693 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1694 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1695 rt->rt6i_metric = 0;
1697 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1698 #ifdef CONFIG_IPV6_SUBTREES
1699 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1701 rt->rt6i_table = ort->rt6i_table;
1706 #ifdef CONFIG_IPV6_ROUTE_INFO
1707 static struct rt6_info *rt6_get_route_info(struct net *net,
1708 struct in6_addr *prefix, int prefixlen,
1709 struct in6_addr *gwaddr, int ifindex)
1711 struct fib6_node *fn;
1712 struct rt6_info *rt = NULL;
1713 struct fib6_table *table;
1715 table = fib6_get_table(net, RT6_TABLE_INFO);
1719 write_lock_bh(&table->tb6_lock);
1720 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1724 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1725 if (rt->rt6i_dev->ifindex != ifindex)
1727 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1729 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1735 write_unlock_bh(&table->tb6_lock);
1739 static struct rt6_info *rt6_add_route_info(struct net *net,
1740 struct in6_addr *prefix, int prefixlen,
1741 struct in6_addr *gwaddr, int ifindex,
1744 struct fib6_config cfg = {
1745 .fc_table = RT6_TABLE_INFO,
1746 .fc_metric = IP6_RT_PRIO_USER,
1747 .fc_ifindex = ifindex,
1748 .fc_dst_len = prefixlen,
1749 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1750 RTF_UP | RTF_PREF(pref),
1752 .fc_nlinfo.nlh = NULL,
1753 .fc_nlinfo.nl_net = net,
1756 ipv6_addr_copy(&cfg.fc_dst, prefix);
1757 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1759 /* We should treat it as a default route if prefix length is 0. */
1761 cfg.fc_flags |= RTF_DEFAULT;
1763 ip6_route_add(&cfg);
1765 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1769 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1771 struct rt6_info *rt;
1772 struct fib6_table *table;
1774 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1778 write_lock_bh(&table->tb6_lock);
1779 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1780 if (dev == rt->rt6i_dev &&
1781 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1782 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1787 write_unlock_bh(&table->tb6_lock);
1791 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1792 struct net_device *dev,
1795 struct fib6_config cfg = {
1796 .fc_table = RT6_TABLE_DFLT,
1797 .fc_metric = IP6_RT_PRIO_USER,
1798 .fc_ifindex = dev->ifindex,
1799 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1800 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1802 .fc_nlinfo.nlh = NULL,
1803 .fc_nlinfo.nl_net = dev_net(dev),
1806 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1808 ip6_route_add(&cfg);
1810 return rt6_get_dflt_router(gwaddr, dev);
1813 void rt6_purge_dflt_routers(struct net *net)
1815 struct rt6_info *rt;
1816 struct fib6_table *table;
1818 /* NOTE: Keep consistent with rt6_get_dflt_router */
1819 table = fib6_get_table(net, RT6_TABLE_DFLT);
1824 read_lock_bh(&table->tb6_lock);
1825 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1826 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1828 read_unlock_bh(&table->tb6_lock);
1833 read_unlock_bh(&table->tb6_lock);
1836 static void rtmsg_to_fib6_config(struct net *net,
1837 struct in6_rtmsg *rtmsg,
1838 struct fib6_config *cfg)
1840 memset(cfg, 0, sizeof(*cfg));
1842 cfg->fc_table = RT6_TABLE_MAIN;
1843 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1844 cfg->fc_metric = rtmsg->rtmsg_metric;
1845 cfg->fc_expires = rtmsg->rtmsg_info;
1846 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1847 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1848 cfg->fc_flags = rtmsg->rtmsg_flags;
1850 cfg->fc_nlinfo.nl_net = net;
1852 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1853 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1854 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1857 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1859 struct fib6_config cfg;
1860 struct in6_rtmsg rtmsg;
1864 case SIOCADDRT: /* Add a route */
1865 case SIOCDELRT: /* Delete a route */
1866 if (!capable(CAP_NET_ADMIN))
1868 err = copy_from_user(&rtmsg, arg,
1869 sizeof(struct in6_rtmsg));
1873 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1878 err = ip6_route_add(&cfg);
1881 err = ip6_route_del(&cfg);
1895 * Drop the packet on the floor
1898 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1901 struct dst_entry *dst = skb_dst(skb);
1902 switch (ipstats_mib_noroutes) {
1903 case IPSTATS_MIB_INNOROUTES:
1904 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1905 if (type == IPV6_ADDR_ANY) {
1906 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1907 IPSTATS_MIB_INADDRERRORS);
1911 case IPSTATS_MIB_OUTNOROUTES:
1912 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1913 ipstats_mib_noroutes);
1916 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1921 static int ip6_pkt_discard(struct sk_buff *skb)
1923 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1926 static int ip6_pkt_discard_out(struct sk_buff *skb)
1928 skb->dev = skb_dst(skb)->dev;
1929 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1932 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1934 static int ip6_pkt_prohibit(struct sk_buff *skb)
1936 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1939 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1941 skb->dev = skb_dst(skb)->dev;
1942 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1948 * Allocate a dst for local (unicast / anycast) address.
1951 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1952 const struct in6_addr *addr,
1955 struct net *net = dev_net(idev->dev);
1956 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1957 struct neighbour *neigh;
1960 if (net_ratelimit())
1961 pr_warning("IPv6: Maximum number of routes reached,"
1962 " consider increasing route/max_size.\n");
1963 return ERR_PTR(-ENOMEM);
1966 dev_hold(net->loopback_dev);
1969 rt->dst.flags = DST_HOST;
1970 rt->dst.input = ip6_input;
1971 rt->dst.output = ip6_output;
1972 rt->rt6i_dev = net->loopback_dev;
1973 rt->rt6i_idev = idev;
1974 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, -1);
1975 rt->dst.obsolete = -1;
1977 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1979 rt->rt6i_flags |= RTF_ANYCAST;
1981 rt->rt6i_flags |= RTF_LOCAL;
1982 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1983 if (IS_ERR(neigh)) {
1986 /* We are casting this because that is the return
1987 * value type. But an errno encoded pointer is the
1988 * same regardless of the underlying pointer type,
1989 * and that's what we are returning. So this is OK.
1991 return (struct rt6_info *) neigh;
1993 rt->rt6i_nexthop = neigh;
1995 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1996 rt->rt6i_dst.plen = 128;
1997 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1999 atomic_set(&rt->dst.__refcnt, 1);
2004 struct arg_dev_net {
2005 struct net_device *dev;
2009 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2011 const struct arg_dev_net *adn = arg;
2012 const struct net_device *dev = adn->dev;
2014 if ((rt->rt6i_dev == dev || dev == NULL) &&
2015 rt != adn->net->ipv6.ip6_null_entry) {
2016 RT6_TRACE("deleted by ifdown %p\n", rt);
2022 void rt6_ifdown(struct net *net, struct net_device *dev)
2024 struct arg_dev_net adn = {
2029 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2030 icmp6_clean_all(fib6_ifdown, &adn);
2033 struct rt6_mtu_change_arg
2035 struct net_device *dev;
2039 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2041 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2042 struct inet6_dev *idev;
2044 /* In IPv6 pmtu discovery is not optional,
2045 so that RTAX_MTU lock cannot disable it.
2046 We still use this lock to block changes
2047 caused by addrconf/ndisc.
2050 idev = __in6_dev_get(arg->dev);
2054 /* For administrative MTU increase, there is no way to discover
2055 IPv6 PMTU increase, so PMTU increase should be updated here.
2056 Since RFC 1981 doesn't include administrative MTU increase
2057 update PMTU increase is a MUST. (i.e. jumbo frame)
2060 If new MTU is less than route PMTU, this new MTU will be the
2061 lowest MTU in the path, update the route PMTU to reflect PMTU
2062 decreases; if new MTU is greater than route PMTU, and the
2063 old MTU is the lowest MTU in the path, update the route PMTU
2064 to reflect the increase. In this case if the other nodes' MTU
2065 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2068 if (rt->rt6i_dev == arg->dev &&
2069 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2070 (dst_mtu(&rt->dst) >= arg->mtu ||
2071 (dst_mtu(&rt->dst) < arg->mtu &&
2072 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2073 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2078 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2080 struct rt6_mtu_change_arg arg = {
2085 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2088 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2089 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2090 [RTA_OIF] = { .type = NLA_U32 },
2091 [RTA_IIF] = { .type = NLA_U32 },
2092 [RTA_PRIORITY] = { .type = NLA_U32 },
2093 [RTA_METRICS] = { .type = NLA_NESTED },
2096 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2097 struct fib6_config *cfg)
2100 struct nlattr *tb[RTA_MAX+1];
2103 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2108 rtm = nlmsg_data(nlh);
2109 memset(cfg, 0, sizeof(*cfg));
2111 cfg->fc_table = rtm->rtm_table;
2112 cfg->fc_dst_len = rtm->rtm_dst_len;
2113 cfg->fc_src_len = rtm->rtm_src_len;
2114 cfg->fc_flags = RTF_UP;
2115 cfg->fc_protocol = rtm->rtm_protocol;
2117 if (rtm->rtm_type == RTN_UNREACHABLE)
2118 cfg->fc_flags |= RTF_REJECT;
2120 if (rtm->rtm_type == RTN_LOCAL)
2121 cfg->fc_flags |= RTF_LOCAL;
2123 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2124 cfg->fc_nlinfo.nlh = nlh;
2125 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2127 if (tb[RTA_GATEWAY]) {
2128 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2129 cfg->fc_flags |= RTF_GATEWAY;
2133 int plen = (rtm->rtm_dst_len + 7) >> 3;
2135 if (nla_len(tb[RTA_DST]) < plen)
2138 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2142 int plen = (rtm->rtm_src_len + 7) >> 3;
2144 if (nla_len(tb[RTA_SRC]) < plen)
2147 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2151 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2153 if (tb[RTA_PRIORITY])
2154 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2156 if (tb[RTA_METRICS]) {
2157 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2158 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2162 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2169 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2171 struct fib6_config cfg;
2174 err = rtm_to_fib6_config(skb, nlh, &cfg);
2178 return ip6_route_del(&cfg);
2181 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2183 struct fib6_config cfg;
2186 err = rtm_to_fib6_config(skb, nlh, &cfg);
2190 return ip6_route_add(&cfg);
2193 static inline size_t rt6_nlmsg_size(void)
2195 return NLMSG_ALIGN(sizeof(struct rtmsg))
2196 + nla_total_size(16) /* RTA_SRC */
2197 + nla_total_size(16) /* RTA_DST */
2198 + nla_total_size(16) /* RTA_GATEWAY */
2199 + nla_total_size(16) /* RTA_PREFSRC */
2200 + nla_total_size(4) /* RTA_TABLE */
2201 + nla_total_size(4) /* RTA_IIF */
2202 + nla_total_size(4) /* RTA_OIF */
2203 + nla_total_size(4) /* RTA_PRIORITY */
2204 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2205 + nla_total_size(sizeof(struct rta_cacheinfo));
2208 static int rt6_fill_node(struct net *net,
2209 struct sk_buff *skb, struct rt6_info *rt,
2210 struct in6_addr *dst, struct in6_addr *src,
2211 int iif, int type, u32 pid, u32 seq,
2212 int prefix, int nowait, unsigned int flags)
2215 struct nlmsghdr *nlh;
2219 if (prefix) { /* user wants prefix routes only */
2220 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2221 /* success since this is not a prefix route */
2226 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2230 rtm = nlmsg_data(nlh);
2231 rtm->rtm_family = AF_INET6;
2232 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2233 rtm->rtm_src_len = rt->rt6i_src.plen;
2236 table = rt->rt6i_table->tb6_id;
2238 table = RT6_TABLE_UNSPEC;
2239 rtm->rtm_table = table;
2240 NLA_PUT_U32(skb, RTA_TABLE, table);
2241 if (rt->rt6i_flags&RTF_REJECT)
2242 rtm->rtm_type = RTN_UNREACHABLE;
2243 else if (rt->rt6i_flags&RTF_LOCAL)
2244 rtm->rtm_type = RTN_LOCAL;
2245 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2246 rtm->rtm_type = RTN_LOCAL;
2248 rtm->rtm_type = RTN_UNICAST;
2250 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2251 rtm->rtm_protocol = rt->rt6i_protocol;
2252 if (rt->rt6i_flags&RTF_DYNAMIC)
2253 rtm->rtm_protocol = RTPROT_REDIRECT;
2254 else if (rt->rt6i_flags & RTF_ADDRCONF)
2255 rtm->rtm_protocol = RTPROT_KERNEL;
2256 else if (rt->rt6i_flags&RTF_DEFAULT)
2257 rtm->rtm_protocol = RTPROT_RA;
2259 if (rt->rt6i_flags&RTF_CACHE)
2260 rtm->rtm_flags |= RTM_F_CLONED;
2263 NLA_PUT(skb, RTA_DST, 16, dst);
2264 rtm->rtm_dst_len = 128;
2265 } else if (rtm->rtm_dst_len)
2266 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2267 #ifdef CONFIG_IPV6_SUBTREES
2269 NLA_PUT(skb, RTA_SRC, 16, src);
2270 rtm->rtm_src_len = 128;
2271 } else if (rtm->rtm_src_len)
2272 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2275 #ifdef CONFIG_IPV6_MROUTE
2276 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2277 int err = ip6mr_get_route(net, skb, rtm, nowait);
2282 goto nla_put_failure;
2284 if (err == -EMSGSIZE)
2285 goto nla_put_failure;
2290 NLA_PUT_U32(skb, RTA_IIF, iif);
2292 struct inet6_dev *idev = ip6_dst_idev(&rt->dst);
2293 struct in6_addr saddr_buf;
2294 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2295 dst, 0, &saddr_buf) == 0)
2296 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2299 if (rtnetlink_put_metrics(skb, dst_metrics_ptr(&rt->dst)) < 0)
2300 goto nla_put_failure;
2302 if (rt->dst.neighbour)
2303 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->dst.neighbour->primary_key);
2306 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2308 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2310 if (!(rt->rt6i_flags & RTF_EXPIRES))
2312 else if (rt->rt6i_expires - jiffies < INT_MAX)
2313 expires = rt->rt6i_expires - jiffies;
2317 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2318 expires, rt->dst.error) < 0)
2319 goto nla_put_failure;
2321 return nlmsg_end(skb, nlh);
2324 nlmsg_cancel(skb, nlh);
2328 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2330 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2333 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2334 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2335 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2339 return rt6_fill_node(arg->net,
2340 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2341 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2342 prefix, 0, NLM_F_MULTI);
2345 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2347 struct net *net = sock_net(in_skb->sk);
2348 struct nlattr *tb[RTA_MAX+1];
2349 struct rt6_info *rt;
2350 struct sk_buff *skb;
2355 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2360 memset(&fl, 0, sizeof(fl));
2363 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2366 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2370 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2373 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2377 iif = nla_get_u32(tb[RTA_IIF]);
2380 fl.oif = nla_get_u32(tb[RTA_OIF]);
2383 struct net_device *dev;
2384 dev = __dev_get_by_index(net, iif);
2391 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2397 /* Reserve room for dummy headers, this skb can pass
2398 through good chunk of routing engine.
2400 skb_reset_mac_header(skb);
2401 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2403 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2404 skb_dst_set(skb, &rt->dst);
2406 err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2407 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2408 nlh->nlmsg_seq, 0, 0, 0);
2414 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2419 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2421 struct sk_buff *skb;
2422 struct net *net = info->nl_net;
2427 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2429 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2433 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2434 event, info->pid, seq, 0, 0, 0);
2436 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2437 WARN_ON(err == -EMSGSIZE);
2441 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2442 info->nlh, gfp_any());
2446 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2449 static int ip6_route_dev_notify(struct notifier_block *this,
2450 unsigned long event, void *data)
2452 struct net_device *dev = (struct net_device *)data;
2453 struct net *net = dev_net(dev);
2455 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2456 net->ipv6.ip6_null_entry->dst.dev = dev;
2457 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2458 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2459 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2460 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2461 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2462 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2473 #ifdef CONFIG_PROC_FS
2484 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2486 struct seq_file *m = p_arg;
2488 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2490 #ifdef CONFIG_IPV6_SUBTREES
2491 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2493 seq_puts(m, "00000000000000000000000000000000 00 ");
2496 if (rt->rt6i_nexthop) {
2497 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2499 seq_puts(m, "00000000000000000000000000000000");
2501 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2502 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2503 rt->dst.__use, rt->rt6i_flags,
2504 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2508 static int ipv6_route_show(struct seq_file *m, void *v)
2510 struct net *net = (struct net *)m->private;
2511 fib6_clean_all(net, rt6_info_route, 0, m);
2515 static int ipv6_route_open(struct inode *inode, struct file *file)
2517 return single_open_net(inode, file, ipv6_route_show);
2520 static const struct file_operations ipv6_route_proc_fops = {
2521 .owner = THIS_MODULE,
2522 .open = ipv6_route_open,
2524 .llseek = seq_lseek,
2525 .release = single_release_net,
2528 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2530 struct net *net = (struct net *)seq->private;
2531 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2532 net->ipv6.rt6_stats->fib_nodes,
2533 net->ipv6.rt6_stats->fib_route_nodes,
2534 net->ipv6.rt6_stats->fib_rt_alloc,
2535 net->ipv6.rt6_stats->fib_rt_entries,
2536 net->ipv6.rt6_stats->fib_rt_cache,
2537 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
2538 net->ipv6.rt6_stats->fib_discarded_routes);
2543 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2545 return single_open_net(inode, file, rt6_stats_seq_show);
2548 static const struct file_operations rt6_stats_seq_fops = {
2549 .owner = THIS_MODULE,
2550 .open = rt6_stats_seq_open,
2552 .llseek = seq_lseek,
2553 .release = single_release_net,
2555 #endif /* CONFIG_PROC_FS */
2557 #ifdef CONFIG_SYSCTL
2560 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2561 void __user *buffer, size_t *lenp, loff_t *ppos)
2563 struct net *net = current->nsproxy->net_ns;
2564 int delay = net->ipv6.sysctl.flush_delay;
2566 proc_dointvec(ctl, write, buffer, lenp, ppos);
2567 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2573 ctl_table ipv6_route_table_template[] = {
2575 .procname = "flush",
2576 .data = &init_net.ipv6.sysctl.flush_delay,
2577 .maxlen = sizeof(int),
2579 .proc_handler = ipv6_sysctl_rtcache_flush
2582 .procname = "gc_thresh",
2583 .data = &ip6_dst_ops_template.gc_thresh,
2584 .maxlen = sizeof(int),
2586 .proc_handler = proc_dointvec,
2589 .procname = "max_size",
2590 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2591 .maxlen = sizeof(int),
2593 .proc_handler = proc_dointvec,
2596 .procname = "gc_min_interval",
2597 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2598 .maxlen = sizeof(int),
2600 .proc_handler = proc_dointvec_jiffies,
2603 .procname = "gc_timeout",
2604 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2605 .maxlen = sizeof(int),
2607 .proc_handler = proc_dointvec_jiffies,
2610 .procname = "gc_interval",
2611 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2612 .maxlen = sizeof(int),
2614 .proc_handler = proc_dointvec_jiffies,
2617 .procname = "gc_elasticity",
2618 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2619 .maxlen = sizeof(int),
2621 .proc_handler = proc_dointvec,
2624 .procname = "mtu_expires",
2625 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2626 .maxlen = sizeof(int),
2628 .proc_handler = proc_dointvec_jiffies,
2631 .procname = "min_adv_mss",
2632 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2633 .maxlen = sizeof(int),
2635 .proc_handler = proc_dointvec,
2638 .procname = "gc_min_interval_ms",
2639 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2640 .maxlen = sizeof(int),
2642 .proc_handler = proc_dointvec_ms_jiffies,
2647 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2649 struct ctl_table *table;
2651 table = kmemdup(ipv6_route_table_template,
2652 sizeof(ipv6_route_table_template),
2656 table[0].data = &net->ipv6.sysctl.flush_delay;
2657 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2658 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2659 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2660 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2661 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2662 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2663 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2664 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2665 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2672 static int __net_init ip6_route_net_init(struct net *net)
2676 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2677 sizeof(net->ipv6.ip6_dst_ops));
2679 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
2680 goto out_ip6_dst_ops;
2682 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2683 sizeof(*net->ipv6.ip6_null_entry),
2685 if (!net->ipv6.ip6_null_entry)
2686 goto out_ip6_dst_entries;
2687 net->ipv6.ip6_null_entry->dst.path =
2688 (struct dst_entry *)net->ipv6.ip6_null_entry;
2689 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2690 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
2691 ip6_template_metrics, true);
2693 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2694 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2695 sizeof(*net->ipv6.ip6_prohibit_entry),
2697 if (!net->ipv6.ip6_prohibit_entry)
2698 goto out_ip6_null_entry;
2699 net->ipv6.ip6_prohibit_entry->dst.path =
2700 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2701 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2702 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
2703 ip6_template_metrics, true);
2705 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2706 sizeof(*net->ipv6.ip6_blk_hole_entry),
2708 if (!net->ipv6.ip6_blk_hole_entry)
2709 goto out_ip6_prohibit_entry;
2710 net->ipv6.ip6_blk_hole_entry->dst.path =
2711 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2712 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2713 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
2714 ip6_template_metrics, true);
2717 net->ipv6.sysctl.flush_delay = 0;
2718 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2719 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2720 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2721 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2722 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2723 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2724 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2726 #ifdef CONFIG_PROC_FS
2727 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2728 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2730 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2736 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2737 out_ip6_prohibit_entry:
2738 kfree(net->ipv6.ip6_prohibit_entry);
2740 kfree(net->ipv6.ip6_null_entry);
2742 out_ip6_dst_entries:
2743 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2748 static void __net_exit ip6_route_net_exit(struct net *net)
2750 #ifdef CONFIG_PROC_FS
2751 proc_net_remove(net, "ipv6_route");
2752 proc_net_remove(net, "rt6_stats");
2754 kfree(net->ipv6.ip6_null_entry);
2755 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2756 kfree(net->ipv6.ip6_prohibit_entry);
2757 kfree(net->ipv6.ip6_blk_hole_entry);
2759 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
2762 static struct pernet_operations ip6_route_net_ops = {
2763 .init = ip6_route_net_init,
2764 .exit = ip6_route_net_exit,
2767 static struct notifier_block ip6_route_dev_notifier = {
2768 .notifier_call = ip6_route_dev_notify,
2772 int __init ip6_route_init(void)
2777 ip6_dst_ops_template.kmem_cachep =
2778 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2779 SLAB_HWCACHE_ALIGN, NULL);
2780 if (!ip6_dst_ops_template.kmem_cachep)
2783 ret = dst_entries_init(&ip6_dst_blackhole_ops);
2785 goto out_kmem_cache;
2787 ret = register_pernet_subsys(&ip6_route_net_ops);
2789 goto out_dst_entries;
2791 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2793 /* Registering of the loopback is done before this portion of code,
2794 * the loopback reference in rt6_info will not be taken, do it
2795 * manually for init_net */
2796 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2797 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2798 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2799 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2800 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2801 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2802 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2806 goto out_register_subsys;
2812 ret = fib6_rules_init();
2817 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2818 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2819 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2820 goto fib6_rules_init;
2822 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2824 goto fib6_rules_init;
2830 fib6_rules_cleanup();
2835 out_register_subsys:
2836 unregister_pernet_subsys(&ip6_route_net_ops);
2838 dst_entries_destroy(&ip6_dst_blackhole_ops);
2840 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2844 void ip6_route_cleanup(void)
2846 unregister_netdevice_notifier(&ip6_route_dev_notifier);
2847 fib6_rules_cleanup();
2850 unregister_pernet_subsys(&ip6_route_net_ops);
2851 dst_entries_destroy(&ip6_dst_blackhole_ops);
2852 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);