2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
18 * YOSHIFUJI Hideaki @USAGI
19 * reworked default router selection.
20 * - respect outgoing interface
21 * - select from (probably) reachable routers (i.e.
22 * routers in REACHABLE, STALE, DELAY or PROBE states).
23 * - always select the same router if it is (probably)
24 * reachable. otherwise, round-robin the list.
26 * Fixed routing subtrees.
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/times.h>
33 #include <linux/socket.h>
34 #include <linux/sockios.h>
35 #include <linux/net.h>
36 #include <linux/route.h>
37 #include <linux/netdevice.h>
38 #include <linux/in6.h>
39 #include <linux/init.h>
40 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <net/net_namespace.h>
46 #include <net/ip6_fib.h>
47 #include <net/ip6_route.h>
48 #include <net/ndisc.h>
49 #include <net/addrconf.h>
51 #include <linux/rtnetlink.h>
54 #include <net/netevent.h>
55 #include <net/netlink.h>
57 #include <asm/uaccess.h>
60 #include <linux/sysctl.h>
63 /* Set to 3 to get tracing. */
67 #define RDBG(x) printk x
68 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
71 #define RT6_TRACE(x...) do { ; } while (0)
74 #define CLONE_OFFLINK_ROUTE 0
76 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
78 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
79 static void ip6_dst_destroy(struct dst_entry *);
80 static void ip6_dst_ifdown(struct dst_entry *,
81 struct net_device *dev, int how);
82 static int ip6_dst_gc(struct dst_ops *ops);
84 static int ip6_pkt_discard(struct sk_buff *skb);
85 static int ip6_pkt_discard_out(struct sk_buff *skb);
86 static void ip6_link_failure(struct sk_buff *skb);
87 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
89 #ifdef CONFIG_IPV6_ROUTE_INFO
90 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
91 struct in6_addr *gwaddr, int ifindex,
93 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
94 struct in6_addr *gwaddr, int ifindex);
97 static struct dst_ops ip6_dst_ops = {
99 .protocol = __constant_htons(ETH_P_IPV6),
102 .check = ip6_dst_check,
103 .destroy = ip6_dst_destroy,
104 .ifdown = ip6_dst_ifdown,
105 .negative_advice = ip6_negative_advice,
106 .link_failure = ip6_link_failure,
107 .update_pmtu = ip6_rt_update_pmtu,
108 .local_out = ip6_local_out,
109 .entry_size = sizeof(struct rt6_info),
110 .entries = ATOMIC_INIT(0),
113 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
117 static struct dst_ops ip6_dst_blackhole_ops = {
119 .protocol = __constant_htons(ETH_P_IPV6),
120 .destroy = ip6_dst_destroy,
121 .check = ip6_dst_check,
122 .update_pmtu = ip6_rt_blackhole_update_pmtu,
123 .entry_size = sizeof(struct rt6_info),
124 .entries = ATOMIC_INIT(0),
127 struct rt6_info ip6_null_entry = {
130 .__refcnt = ATOMIC_INIT(1),
133 .error = -ENETUNREACH,
134 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
135 .input = ip6_pkt_discard,
136 .output = ip6_pkt_discard_out,
138 .path = (struct dst_entry*)&ip6_null_entry,
141 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
142 .rt6i_metric = ~(u32) 0,
143 .rt6i_ref = ATOMIC_INIT(1),
146 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
148 static int ip6_pkt_prohibit(struct sk_buff *skb);
149 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
151 struct rt6_info ip6_prohibit_entry = {
154 .__refcnt = ATOMIC_INIT(1),
158 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
159 .input = ip6_pkt_prohibit,
160 .output = ip6_pkt_prohibit_out,
162 .path = (struct dst_entry*)&ip6_prohibit_entry,
165 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
166 .rt6i_metric = ~(u32) 0,
167 .rt6i_ref = ATOMIC_INIT(1),
170 struct rt6_info ip6_blk_hole_entry = {
173 .__refcnt = ATOMIC_INIT(1),
177 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
178 .input = dst_discard,
179 .output = dst_discard,
181 .path = (struct dst_entry*)&ip6_blk_hole_entry,
184 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
185 .rt6i_metric = ~(u32) 0,
186 .rt6i_ref = ATOMIC_INIT(1),
191 /* allocate dst with ip6_dst_ops */
192 static __inline__ struct rt6_info *ip6_dst_alloc(void)
194 return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
197 static void ip6_dst_destroy(struct dst_entry *dst)
199 struct rt6_info *rt = (struct rt6_info *)dst;
200 struct inet6_dev *idev = rt->rt6i_idev;
203 rt->rt6i_idev = NULL;
208 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
211 struct rt6_info *rt = (struct rt6_info *)dst;
212 struct inet6_dev *idev = rt->rt6i_idev;
213 struct net_device *loopback_dev =
214 dev->nd_net->loopback_dev;
216 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
217 struct inet6_dev *loopback_idev =
218 in6_dev_get(loopback_dev);
219 if (loopback_idev != NULL) {
220 rt->rt6i_idev = loopback_idev;
226 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
228 return (rt->rt6i_flags & RTF_EXPIRES &&
229 time_after(jiffies, rt->rt6i_expires));
232 static inline int rt6_need_strict(struct in6_addr *daddr)
234 return (ipv6_addr_type(daddr) &
235 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
239 * Route lookup. Any table->tb6_lock is implied.
242 static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
246 struct rt6_info *local = NULL;
247 struct rt6_info *sprt;
250 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
251 struct net_device *dev = sprt->rt6i_dev;
252 if (dev->ifindex == oif)
254 if (dev->flags & IFF_LOOPBACK) {
255 if (sprt->rt6i_idev == NULL ||
256 sprt->rt6i_idev->dev->ifindex != oif) {
259 if (local && (!oif ||
260 local->rt6i_idev->dev->ifindex == oif))
271 return &ip6_null_entry;
276 #ifdef CONFIG_IPV6_ROUTER_PREF
277 static void rt6_probe(struct rt6_info *rt)
279 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
281 * Okay, this does not seem to be appropriate
282 * for now, however, we need to check if it
283 * is really so; aka Router Reachability Probing.
285 * Router Reachability Probe MUST be rate-limited
286 * to no more than one per minute.
288 if (!neigh || (neigh->nud_state & NUD_VALID))
290 read_lock_bh(&neigh->lock);
291 if (!(neigh->nud_state & NUD_VALID) &&
292 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
293 struct in6_addr mcaddr;
294 struct in6_addr *target;
296 neigh->updated = jiffies;
297 read_unlock_bh(&neigh->lock);
299 target = (struct in6_addr *)&neigh->primary_key;
300 addrconf_addr_solict_mult(target, &mcaddr);
301 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
303 read_unlock_bh(&neigh->lock);
306 static inline void rt6_probe(struct rt6_info *rt)
313 * Default Router Selection (RFC 2461 6.3.6)
315 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
317 struct net_device *dev = rt->rt6i_dev;
318 if (!oif || dev->ifindex == oif)
320 if ((dev->flags & IFF_LOOPBACK) &&
321 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
326 static inline int rt6_check_neigh(struct rt6_info *rt)
328 struct neighbour *neigh = rt->rt6i_nexthop;
330 if (rt->rt6i_flags & RTF_NONEXTHOP ||
331 !(rt->rt6i_flags & RTF_GATEWAY))
334 read_lock_bh(&neigh->lock);
335 if (neigh->nud_state & NUD_VALID)
337 #ifdef CONFIG_IPV6_ROUTER_PREF
338 else if (neigh->nud_state & NUD_FAILED)
343 read_unlock_bh(&neigh->lock);
349 static int rt6_score_route(struct rt6_info *rt, int oif,
354 m = rt6_check_dev(rt, oif);
355 if (!m && (strict & RT6_LOOKUP_F_IFACE))
357 #ifdef CONFIG_IPV6_ROUTER_PREF
358 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
360 n = rt6_check_neigh(rt);
361 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
366 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
367 int *mpri, struct rt6_info *match)
371 if (rt6_check_expired(rt))
374 m = rt6_score_route(rt, oif, strict);
379 if (strict & RT6_LOOKUP_F_REACHABLE)
383 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
391 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
392 struct rt6_info *rr_head,
393 u32 metric, int oif, int strict)
395 struct rt6_info *rt, *match;
399 for (rt = rr_head; rt && rt->rt6i_metric == metric;
400 rt = rt->u.dst.rt6_next)
401 match = find_match(rt, oif, strict, &mpri, match);
402 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
403 rt = rt->u.dst.rt6_next)
404 match = find_match(rt, oif, strict, &mpri, match);
409 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
411 struct rt6_info *match, *rt0;
413 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
414 __FUNCTION__, fn->leaf, oif);
418 fn->rr_ptr = rt0 = fn->leaf;
420 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
423 (strict & RT6_LOOKUP_F_REACHABLE)) {
424 struct rt6_info *next = rt0->u.dst.rt6_next;
426 /* no entries matched; do round-robin */
427 if (!next || next->rt6i_metric != rt0->rt6i_metric)
434 RT6_TRACE("%s() => %p\n",
435 __FUNCTION__, match);
437 return (match ? match : &ip6_null_entry);
440 #ifdef CONFIG_IPV6_ROUTE_INFO
441 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
442 struct in6_addr *gwaddr)
444 struct route_info *rinfo = (struct route_info *) opt;
445 struct in6_addr prefix_buf, *prefix;
450 if (len < sizeof(struct route_info)) {
454 /* Sanity check for prefix_len and length */
455 if (rinfo->length > 3) {
457 } else if (rinfo->prefix_len > 128) {
459 } else if (rinfo->prefix_len > 64) {
460 if (rinfo->length < 2) {
463 } else if (rinfo->prefix_len > 0) {
464 if (rinfo->length < 1) {
469 pref = rinfo->route_pref;
470 if (pref == ICMPV6_ROUTER_PREF_INVALID)
471 pref = ICMPV6_ROUTER_PREF_MEDIUM;
473 lifetime = ntohl(rinfo->lifetime);
474 if (lifetime == 0xffffffff) {
476 } else if (lifetime > 0x7fffffff/HZ) {
477 /* Avoid arithmetic overflow */
478 lifetime = 0x7fffffff/HZ - 1;
481 if (rinfo->length == 3)
482 prefix = (struct in6_addr *)rinfo->prefix;
484 /* this function is safe */
485 ipv6_addr_prefix(&prefix_buf,
486 (struct in6_addr *)rinfo->prefix,
488 prefix = &prefix_buf;
491 rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
493 if (rt && !lifetime) {
499 rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
502 rt->rt6i_flags = RTF_ROUTEINFO |
503 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
506 if (lifetime == 0xffffffff) {
507 rt->rt6i_flags &= ~RTF_EXPIRES;
509 rt->rt6i_expires = jiffies + HZ * lifetime;
510 rt->rt6i_flags |= RTF_EXPIRES;
512 dst_release(&rt->u.dst);
518 #define BACKTRACK(saddr) \
520 if (rt == &ip6_null_entry) { \
521 struct fib6_node *pn; \
523 if (fn->fn_flags & RTN_TL_ROOT) \
526 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
527 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
530 if (fn->fn_flags & RTN_RTINFO) \
536 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
537 struct flowi *fl, int flags)
539 struct fib6_node *fn;
542 read_lock_bh(&table->tb6_lock);
543 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
546 rt = rt6_device_match(rt, fl->oif, flags);
547 BACKTRACK(&fl->fl6_src);
549 dst_use(&rt->u.dst, jiffies);
550 read_unlock_bh(&table->tb6_lock);
555 struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
566 struct dst_entry *dst;
567 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
570 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
571 flags |= RT6_LOOKUP_F_HAS_SADDR;
574 dst = fib6_rule_lookup(&init_net, &fl, flags, ip6_pol_route_lookup);
576 return (struct rt6_info *) dst;
583 EXPORT_SYMBOL(rt6_lookup);
585 /* ip6_ins_rt is called with FREE table->tb6_lock.
586 It takes new route entry, the addition fails by any reason the
587 route is freed. In any case, if caller does not hold it, it may
591 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
594 struct fib6_table *table;
596 table = rt->rt6i_table;
597 write_lock_bh(&table->tb6_lock);
598 err = fib6_add(&table->tb6_root, rt, info);
599 write_unlock_bh(&table->tb6_lock);
604 int ip6_ins_rt(struct rt6_info *rt)
606 struct nl_info info = {
609 return __ip6_ins_rt(rt, &info);
612 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
613 struct in6_addr *saddr)
621 rt = ip6_rt_copy(ort);
624 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
625 if (rt->rt6i_dst.plen != 128 &&
626 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
627 rt->rt6i_flags |= RTF_ANYCAST;
628 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
631 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
632 rt->rt6i_dst.plen = 128;
633 rt->rt6i_flags |= RTF_CACHE;
634 rt->u.dst.flags |= DST_HOST;
636 #ifdef CONFIG_IPV6_SUBTREES
637 if (rt->rt6i_src.plen && saddr) {
638 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
639 rt->rt6i_src.plen = 128;
643 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
650 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
652 struct rt6_info *rt = ip6_rt_copy(ort);
654 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
655 rt->rt6i_dst.plen = 128;
656 rt->rt6i_flags |= RTF_CACHE;
657 rt->u.dst.flags |= DST_HOST;
658 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
663 static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
664 struct flowi *fl, int flags)
666 struct fib6_node *fn;
667 struct rt6_info *rt, *nrt;
671 int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
673 strict |= flags & RT6_LOOKUP_F_IFACE;
676 read_lock_bh(&table->tb6_lock);
679 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
682 rt = rt6_select(fn, oif, strict | reachable);
683 BACKTRACK(&fl->fl6_src);
684 if (rt == &ip6_null_entry ||
685 rt->rt6i_flags & RTF_CACHE)
688 dst_hold(&rt->u.dst);
689 read_unlock_bh(&table->tb6_lock);
691 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
692 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
694 #if CLONE_OFFLINK_ROUTE
695 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
701 dst_release(&rt->u.dst);
702 rt = nrt ? : &ip6_null_entry;
704 dst_hold(&rt->u.dst);
706 err = ip6_ins_rt(nrt);
715 * Race condition! In the gap, when table->tb6_lock was
716 * released someone could insert this route. Relookup.
718 dst_release(&rt->u.dst);
726 dst_hold(&rt->u.dst);
727 read_unlock_bh(&table->tb6_lock);
729 rt->u.dst.lastuse = jiffies;
735 static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
736 struct flowi *fl, int flags)
738 return ip6_pol_route(table, fl->iif, fl, flags);
741 void ip6_route_input(struct sk_buff *skb)
743 struct ipv6hdr *iph = ipv6_hdr(skb);
744 int flags = RT6_LOOKUP_F_HAS_SADDR;
746 .iif = skb->dev->ifindex,
751 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
755 .proto = iph->nexthdr,
758 if (rt6_need_strict(&iph->daddr))
759 flags |= RT6_LOOKUP_F_IFACE;
761 skb->dst = fib6_rule_lookup(&init_net, &fl, flags, ip6_pol_route_input);
764 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
765 struct flowi *fl, int flags)
767 return ip6_pol_route(table, fl->oif, fl, flags);
770 struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
774 if (rt6_need_strict(&fl->fl6_dst))
775 flags |= RT6_LOOKUP_F_IFACE;
777 if (!ipv6_addr_any(&fl->fl6_src))
778 flags |= RT6_LOOKUP_F_HAS_SADDR;
780 return fib6_rule_lookup(&init_net, fl, flags, ip6_pol_route_output);
783 EXPORT_SYMBOL(ip6_route_output);
785 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
787 struct rt6_info *ort = (struct rt6_info *) *dstp;
788 struct rt6_info *rt = (struct rt6_info *)
789 dst_alloc(&ip6_dst_blackhole_ops);
790 struct dst_entry *new = NULL;
795 atomic_set(&new->__refcnt, 1);
797 new->input = dst_discard;
798 new->output = dst_discard;
800 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
801 new->dev = ort->u.dst.dev;
804 rt->rt6i_idev = ort->rt6i_idev;
806 in6_dev_hold(rt->rt6i_idev);
807 rt->rt6i_expires = 0;
809 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
810 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
813 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
814 #ifdef CONFIG_IPV6_SUBTREES
815 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
823 return (new ? 0 : -ENOMEM);
825 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
828 * Destination cache support functions
831 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
835 rt = (struct rt6_info *) dst;
837 if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
843 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
845 struct rt6_info *rt = (struct rt6_info *) dst;
848 if (rt->rt6i_flags & RTF_CACHE)
856 static void ip6_link_failure(struct sk_buff *skb)
860 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
862 rt = (struct rt6_info *) skb->dst;
864 if (rt->rt6i_flags&RTF_CACHE) {
865 dst_set_expires(&rt->u.dst, 0);
866 rt->rt6i_flags |= RTF_EXPIRES;
867 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
868 rt->rt6i_node->fn_sernum = -1;
872 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
874 struct rt6_info *rt6 = (struct rt6_info*)dst;
876 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
877 rt6->rt6i_flags |= RTF_MODIFIED;
878 if (mtu < IPV6_MIN_MTU) {
880 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
882 dst->metrics[RTAX_MTU-1] = mtu;
883 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
887 static int ipv6_get_mtu(struct net_device *dev);
889 static inline unsigned int ipv6_advmss(unsigned int mtu)
891 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
893 if (mtu < init_net.ipv6.sysctl.ip6_rt_min_advmss)
894 mtu = init_net.ipv6.sysctl.ip6_rt_min_advmss;
897 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
898 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
899 * IPV6_MAXPLEN is also valid and means: "any MSS,
900 * rely only on pmtu discovery"
902 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
907 static struct dst_entry *icmp6_dst_gc_list;
908 static DEFINE_SPINLOCK(icmp6_dst_lock);
910 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
911 struct neighbour *neigh,
912 struct in6_addr *addr)
915 struct inet6_dev *idev = in6_dev_get(dev);
917 if (unlikely(idev == NULL))
920 rt = ip6_dst_alloc();
921 if (unlikely(rt == NULL)) {
930 neigh = ndisc_get_neigh(dev, addr);
933 rt->rt6i_idev = idev;
934 rt->rt6i_nexthop = neigh;
935 atomic_set(&rt->u.dst.__refcnt, 1);
936 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
937 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
938 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
939 rt->u.dst.output = ip6_output;
941 #if 0 /* there's no chance to use these for ndisc */
942 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
945 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
946 rt->rt6i_dst.plen = 128;
949 spin_lock_bh(&icmp6_dst_lock);
950 rt->u.dst.next = icmp6_dst_gc_list;
951 icmp6_dst_gc_list = &rt->u.dst;
952 spin_unlock_bh(&icmp6_dst_lock);
954 fib6_force_start_gc();
960 int icmp6_dst_gc(int *more)
962 struct dst_entry *dst, *next, **pprev;
968 spin_lock_bh(&icmp6_dst_lock);
969 pprev = &icmp6_dst_gc_list;
971 while ((dst = *pprev) != NULL) {
972 if (!atomic_read(&dst->__refcnt)) {
982 spin_unlock_bh(&icmp6_dst_lock);
987 static int ip6_dst_gc(struct dst_ops *ops)
989 static unsigned expire = 30*HZ;
990 static unsigned long last_gc;
991 unsigned long now = jiffies;
993 if (time_after(last_gc + init_net.ipv6.sysctl.ip6_rt_gc_min_interval, now) &&
994 atomic_read(&ip6_dst_ops.entries) <= init_net.ipv6.sysctl.ip6_rt_max_size)
1000 if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
1001 expire = init_net.ipv6.sysctl.ip6_rt_gc_timeout>>1;
1004 expire -= expire>>init_net.ipv6.sysctl.ip6_rt_gc_elasticity;
1005 return (atomic_read(&ip6_dst_ops.entries) > init_net.ipv6.sysctl.ip6_rt_max_size);
1008 /* Clean host part of a prefix. Not necessary in radix tree,
1009 but results in cleaner routing tables.
1011 Remove it only when all the things will work!
1014 static int ipv6_get_mtu(struct net_device *dev)
1016 int mtu = IPV6_MIN_MTU;
1017 struct inet6_dev *idev;
1019 idev = in6_dev_get(dev);
1021 mtu = idev->cnf.mtu6;
1027 int ipv6_get_hoplimit(struct net_device *dev)
1029 int hoplimit = ipv6_devconf.hop_limit;
1030 struct inet6_dev *idev;
1032 idev = in6_dev_get(dev);
1034 hoplimit = idev->cnf.hop_limit;
1044 int ip6_route_add(struct fib6_config *cfg)
1047 struct rt6_info *rt = NULL;
1048 struct net_device *dev = NULL;
1049 struct inet6_dev *idev = NULL;
1050 struct fib6_table *table;
1053 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1055 #ifndef CONFIG_IPV6_SUBTREES
1056 if (cfg->fc_src_len)
1059 if (cfg->fc_ifindex) {
1061 dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
1064 idev = in6_dev_get(dev);
1069 if (cfg->fc_metric == 0)
1070 cfg->fc_metric = IP6_RT_PRIO_USER;
1072 table = fib6_new_table(&init_net, cfg->fc_table);
1073 if (table == NULL) {
1078 rt = ip6_dst_alloc();
1085 rt->u.dst.obsolete = -1;
1086 rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1088 if (cfg->fc_protocol == RTPROT_UNSPEC)
1089 cfg->fc_protocol = RTPROT_BOOT;
1090 rt->rt6i_protocol = cfg->fc_protocol;
1092 addr_type = ipv6_addr_type(&cfg->fc_dst);
1094 if (addr_type & IPV6_ADDR_MULTICAST)
1095 rt->u.dst.input = ip6_mc_input;
1097 rt->u.dst.input = ip6_forward;
1099 rt->u.dst.output = ip6_output;
1101 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1102 rt->rt6i_dst.plen = cfg->fc_dst_len;
1103 if (rt->rt6i_dst.plen == 128)
1104 rt->u.dst.flags = DST_HOST;
1106 #ifdef CONFIG_IPV6_SUBTREES
1107 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1108 rt->rt6i_src.plen = cfg->fc_src_len;
1111 rt->rt6i_metric = cfg->fc_metric;
1113 /* We cannot add true routes via loopback here,
1114 they would result in kernel looping; promote them to reject routes
1116 if ((cfg->fc_flags & RTF_REJECT) ||
1117 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1118 /* hold loopback dev/idev if we haven't done so. */
1119 if (dev != init_net.loopback_dev) {
1124 dev = init_net.loopback_dev;
1126 idev = in6_dev_get(dev);
1132 rt->u.dst.output = ip6_pkt_discard_out;
1133 rt->u.dst.input = ip6_pkt_discard;
1134 rt->u.dst.error = -ENETUNREACH;
1135 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1139 if (cfg->fc_flags & RTF_GATEWAY) {
1140 struct in6_addr *gw_addr;
1143 gw_addr = &cfg->fc_gateway;
1144 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1145 gwa_type = ipv6_addr_type(gw_addr);
1147 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1148 struct rt6_info *grt;
1150 /* IPv6 strictly inhibits using not link-local
1151 addresses as nexthop address.
1152 Otherwise, router will not able to send redirects.
1153 It is very good, but in some (rare!) circumstances
1154 (SIT, PtP, NBMA NOARP links) it is handy to allow
1155 some exceptions. --ANK
1158 if (!(gwa_type&IPV6_ADDR_UNICAST))
1161 grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1163 err = -EHOSTUNREACH;
1167 if (dev != grt->rt6i_dev) {
1168 dst_release(&grt->u.dst);
1172 dev = grt->rt6i_dev;
1173 idev = grt->rt6i_idev;
1175 in6_dev_hold(grt->rt6i_idev);
1177 if (!(grt->rt6i_flags&RTF_GATEWAY))
1179 dst_release(&grt->u.dst);
1185 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1193 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1194 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1195 if (IS_ERR(rt->rt6i_nexthop)) {
1196 err = PTR_ERR(rt->rt6i_nexthop);
1197 rt->rt6i_nexthop = NULL;
1202 rt->rt6i_flags = cfg->fc_flags;
1209 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1210 int type = nla_type(nla);
1213 if (type > RTAX_MAX) {
1218 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1223 if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1224 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1225 if (!rt->u.dst.metrics[RTAX_MTU-1])
1226 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1227 if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1228 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1229 rt->u.dst.dev = dev;
1230 rt->rt6i_idev = idev;
1231 rt->rt6i_table = table;
1232 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1240 dst_free(&rt->u.dst);
1244 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1247 struct fib6_table *table;
1249 if (rt == &ip6_null_entry)
1252 table = rt->rt6i_table;
1253 write_lock_bh(&table->tb6_lock);
1255 err = fib6_del(rt, info);
1256 dst_release(&rt->u.dst);
1258 write_unlock_bh(&table->tb6_lock);
1263 int ip6_del_rt(struct rt6_info *rt)
1265 struct nl_info info = {
1266 .nl_net = &init_net,
1268 return __ip6_del_rt(rt, &info);
1271 static int ip6_route_del(struct fib6_config *cfg)
1273 struct fib6_table *table;
1274 struct fib6_node *fn;
1275 struct rt6_info *rt;
1278 table = fib6_get_table(&init_net, cfg->fc_table);
1282 read_lock_bh(&table->tb6_lock);
1284 fn = fib6_locate(&table->tb6_root,
1285 &cfg->fc_dst, cfg->fc_dst_len,
1286 &cfg->fc_src, cfg->fc_src_len);
1289 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1290 if (cfg->fc_ifindex &&
1291 (rt->rt6i_dev == NULL ||
1292 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1294 if (cfg->fc_flags & RTF_GATEWAY &&
1295 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1297 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1299 dst_hold(&rt->u.dst);
1300 read_unlock_bh(&table->tb6_lock);
1302 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1305 read_unlock_bh(&table->tb6_lock);
1313 struct ip6rd_flowi {
1315 struct in6_addr gateway;
1318 static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1322 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1323 struct rt6_info *rt;
1324 struct fib6_node *fn;
1327 * Get the "current" route for this destination and
1328 * check if the redirect has come from approriate router.
1330 * RFC 2461 specifies that redirects should only be
1331 * accepted if they come from the nexthop to the target.
1332 * Due to the way the routes are chosen, this notion
1333 * is a bit fuzzy and one might need to check all possible
1337 read_lock_bh(&table->tb6_lock);
1338 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1340 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1342 * Current route is on-link; redirect is always invalid.
1344 * Seems, previous statement is not true. It could
1345 * be node, which looks for us as on-link (f.e. proxy ndisc)
1346 * But then router serving it might decide, that we should
1347 * know truth 8)8) --ANK (980726).
1349 if (rt6_check_expired(rt))
1351 if (!(rt->rt6i_flags & RTF_GATEWAY))
1353 if (fl->oif != rt->rt6i_dev->ifindex)
1355 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1361 rt = &ip6_null_entry;
1362 BACKTRACK(&fl->fl6_src);
1364 dst_hold(&rt->u.dst);
1366 read_unlock_bh(&table->tb6_lock);
1371 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1372 struct in6_addr *src,
1373 struct in6_addr *gateway,
1374 struct net_device *dev)
1376 int flags = RT6_LOOKUP_F_HAS_SADDR;
1377 struct ip6rd_flowi rdfl = {
1379 .oif = dev->ifindex,
1387 .gateway = *gateway,
1390 if (rt6_need_strict(dest))
1391 flags |= RT6_LOOKUP_F_IFACE;
1393 return (struct rt6_info *)fib6_rule_lookup(&init_net,
1394 (struct flowi *)&rdfl,
1395 flags, __ip6_route_redirect);
1398 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1399 struct in6_addr *saddr,
1400 struct neighbour *neigh, u8 *lladdr, int on_link)
1402 struct rt6_info *rt, *nrt = NULL;
1403 struct netevent_redirect netevent;
1405 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1407 if (rt == &ip6_null_entry) {
1408 if (net_ratelimit())
1409 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1410 "for redirect target\n");
1415 * We have finally decided to accept it.
1418 neigh_update(neigh, lladdr, NUD_STALE,
1419 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1420 NEIGH_UPDATE_F_OVERRIDE|
1421 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1422 NEIGH_UPDATE_F_ISROUTER))
1426 * Redirect received -> path was valid.
1427 * Look, redirects are sent only in response to data packets,
1428 * so that this nexthop apparently is reachable. --ANK
1430 dst_confirm(&rt->u.dst);
1432 /* Duplicate redirect: silently ignore. */
1433 if (neigh == rt->u.dst.neighbour)
1436 nrt = ip6_rt_copy(rt);
1440 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1442 nrt->rt6i_flags &= ~RTF_GATEWAY;
1444 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1445 nrt->rt6i_dst.plen = 128;
1446 nrt->u.dst.flags |= DST_HOST;
1448 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1449 nrt->rt6i_nexthop = neigh_clone(neigh);
1450 /* Reset pmtu, it may be better */
1451 nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1452 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
1454 if (ip6_ins_rt(nrt))
1457 netevent.old = &rt->u.dst;
1458 netevent.new = &nrt->u.dst;
1459 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1461 if (rt->rt6i_flags&RTF_CACHE) {
1467 dst_release(&rt->u.dst);
1472 * Handle ICMP "packet too big" messages
1473 * i.e. Path MTU discovery
1476 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1477 struct net_device *dev, u32 pmtu)
1479 struct rt6_info *rt, *nrt;
1482 rt = rt6_lookup(daddr, saddr, dev->ifindex, 0);
1486 if (pmtu >= dst_mtu(&rt->u.dst))
1489 if (pmtu < IPV6_MIN_MTU) {
1491 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1492 * MTU (1280) and a fragment header should always be included
1493 * after a node receiving Too Big message reporting PMTU is
1494 * less than the IPv6 Minimum Link MTU.
1496 pmtu = IPV6_MIN_MTU;
1500 /* New mtu received -> path was valid.
1501 They are sent only in response to data packets,
1502 so that this nexthop apparently is reachable. --ANK
1504 dst_confirm(&rt->u.dst);
1506 /* Host route. If it is static, it would be better
1507 not to override it, but add new one, so that
1508 when cache entry will expire old pmtu
1509 would return automatically.
1511 if (rt->rt6i_flags & RTF_CACHE) {
1512 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1514 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1515 dst_set_expires(&rt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1516 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1521 Two cases are possible:
1522 1. It is connected route. Action: COW
1523 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1525 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1526 nrt = rt6_alloc_cow(rt, daddr, saddr);
1528 nrt = rt6_alloc_clone(rt, daddr);
1531 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1533 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1535 /* According to RFC 1981, detecting PMTU increase shouldn't be
1536 * happened within 5 mins, the recommended timer is 10 mins.
1537 * Here this route expiration time is set to ip6_rt_mtu_expires
1538 * which is 10 mins. After 10 mins the decreased pmtu is expired
1539 * and detecting PMTU increase will be automatically happened.
1541 dst_set_expires(&nrt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1542 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1547 dst_release(&rt->u.dst);
1551 * Misc support functions
1554 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1556 struct rt6_info *rt = ip6_dst_alloc();
1559 rt->u.dst.input = ort->u.dst.input;
1560 rt->u.dst.output = ort->u.dst.output;
1562 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1563 rt->u.dst.error = ort->u.dst.error;
1564 rt->u.dst.dev = ort->u.dst.dev;
1566 dev_hold(rt->u.dst.dev);
1567 rt->rt6i_idev = ort->rt6i_idev;
1569 in6_dev_hold(rt->rt6i_idev);
1570 rt->u.dst.lastuse = jiffies;
1571 rt->rt6i_expires = 0;
1573 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1574 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1575 rt->rt6i_metric = 0;
1577 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1578 #ifdef CONFIG_IPV6_SUBTREES
1579 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1581 rt->rt6i_table = ort->rt6i_table;
1586 #ifdef CONFIG_IPV6_ROUTE_INFO
1587 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
1588 struct in6_addr *gwaddr, int ifindex)
1590 struct fib6_node *fn;
1591 struct rt6_info *rt = NULL;
1592 struct fib6_table *table;
1594 table = fib6_get_table(&init_net, RT6_TABLE_INFO);
1598 write_lock_bh(&table->tb6_lock);
1599 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1603 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1604 if (rt->rt6i_dev->ifindex != ifindex)
1606 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1608 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1610 dst_hold(&rt->u.dst);
1614 write_unlock_bh(&table->tb6_lock);
1618 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1619 struct in6_addr *gwaddr, int ifindex,
1622 struct fib6_config cfg = {
1623 .fc_table = RT6_TABLE_INFO,
1624 .fc_metric = IP6_RT_PRIO_USER,
1625 .fc_ifindex = ifindex,
1626 .fc_dst_len = prefixlen,
1627 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1628 RTF_UP | RTF_PREF(pref),
1631 ipv6_addr_copy(&cfg.fc_dst, prefix);
1632 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1634 /* We should treat it as a default route if prefix length is 0. */
1636 cfg.fc_flags |= RTF_DEFAULT;
1638 ip6_route_add(&cfg);
1640 return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1644 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1646 struct rt6_info *rt;
1647 struct fib6_table *table;
1649 table = fib6_get_table(&init_net, RT6_TABLE_DFLT);
1653 write_lock_bh(&table->tb6_lock);
1654 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1655 if (dev == rt->rt6i_dev &&
1656 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1657 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1661 dst_hold(&rt->u.dst);
1662 write_unlock_bh(&table->tb6_lock);
1666 EXPORT_SYMBOL(rt6_get_dflt_router);
1668 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1669 struct net_device *dev,
1672 struct fib6_config cfg = {
1673 .fc_table = RT6_TABLE_DFLT,
1674 .fc_metric = IP6_RT_PRIO_USER,
1675 .fc_ifindex = dev->ifindex,
1676 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1677 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1680 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1682 ip6_route_add(&cfg);
1684 return rt6_get_dflt_router(gwaddr, dev);
1687 void rt6_purge_dflt_routers(void)
1689 struct rt6_info *rt;
1690 struct fib6_table *table;
1692 /* NOTE: Keep consistent with rt6_get_dflt_router */
1693 table = fib6_get_table(&init_net, RT6_TABLE_DFLT);
1698 read_lock_bh(&table->tb6_lock);
1699 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1700 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1701 dst_hold(&rt->u.dst);
1702 read_unlock_bh(&table->tb6_lock);
1707 read_unlock_bh(&table->tb6_lock);
1710 static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1711 struct fib6_config *cfg)
1713 memset(cfg, 0, sizeof(*cfg));
1715 cfg->fc_table = RT6_TABLE_MAIN;
1716 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1717 cfg->fc_metric = rtmsg->rtmsg_metric;
1718 cfg->fc_expires = rtmsg->rtmsg_info;
1719 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1720 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1721 cfg->fc_flags = rtmsg->rtmsg_flags;
1723 cfg->fc_nlinfo.nl_net = &init_net;
1725 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1726 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1727 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1730 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1732 struct fib6_config cfg;
1733 struct in6_rtmsg rtmsg;
1737 case SIOCADDRT: /* Add a route */
1738 case SIOCDELRT: /* Delete a route */
1739 if (!capable(CAP_NET_ADMIN))
1741 err = copy_from_user(&rtmsg, arg,
1742 sizeof(struct in6_rtmsg));
1746 rtmsg_to_fib6_config(&rtmsg, &cfg);
1751 err = ip6_route_add(&cfg);
1754 err = ip6_route_del(&cfg);
1768 * Drop the packet on the floor
1771 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1774 switch (ipstats_mib_noroutes) {
1775 case IPSTATS_MIB_INNOROUTES:
1776 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1777 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1778 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1782 case IPSTATS_MIB_OUTNOROUTES:
1783 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1786 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1791 static int ip6_pkt_discard(struct sk_buff *skb)
1793 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1796 static int ip6_pkt_discard_out(struct sk_buff *skb)
1798 skb->dev = skb->dst->dev;
1799 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1802 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1804 static int ip6_pkt_prohibit(struct sk_buff *skb)
1806 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1809 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1811 skb->dev = skb->dst->dev;
1812 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1818 * Allocate a dst for local (unicast / anycast) address.
1821 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1822 const struct in6_addr *addr,
1825 struct rt6_info *rt = ip6_dst_alloc();
1828 return ERR_PTR(-ENOMEM);
1830 dev_hold(init_net.loopback_dev);
1833 rt->u.dst.flags = DST_HOST;
1834 rt->u.dst.input = ip6_input;
1835 rt->u.dst.output = ip6_output;
1836 rt->rt6i_dev = init_net.loopback_dev;
1837 rt->rt6i_idev = idev;
1838 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1839 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1840 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1841 rt->u.dst.obsolete = -1;
1843 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1845 rt->rt6i_flags |= RTF_ANYCAST;
1847 rt->rt6i_flags |= RTF_LOCAL;
1848 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1849 if (rt->rt6i_nexthop == NULL) {
1850 dst_free(&rt->u.dst);
1851 return ERR_PTR(-ENOMEM);
1854 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1855 rt->rt6i_dst.plen = 128;
1856 rt->rt6i_table = fib6_get_table(&init_net, RT6_TABLE_LOCAL);
1858 atomic_set(&rt->u.dst.__refcnt, 1);
1863 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1865 if (((void*)rt->rt6i_dev == arg || arg == NULL) &&
1866 rt != &ip6_null_entry) {
1867 RT6_TRACE("deleted by ifdown %p\n", rt);
1873 void rt6_ifdown(struct net *net, struct net_device *dev)
1875 fib6_clean_all(net, fib6_ifdown, 0, dev);
1878 struct rt6_mtu_change_arg
1880 struct net_device *dev;
1884 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1886 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1887 struct inet6_dev *idev;
1889 /* In IPv6 pmtu discovery is not optional,
1890 so that RTAX_MTU lock cannot disable it.
1891 We still use this lock to block changes
1892 caused by addrconf/ndisc.
1895 idev = __in6_dev_get(arg->dev);
1899 /* For administrative MTU increase, there is no way to discover
1900 IPv6 PMTU increase, so PMTU increase should be updated here.
1901 Since RFC 1981 doesn't include administrative MTU increase
1902 update PMTU increase is a MUST. (i.e. jumbo frame)
1905 If new MTU is less than route PMTU, this new MTU will be the
1906 lowest MTU in the path, update the route PMTU to reflect PMTU
1907 decreases; if new MTU is greater than route PMTU, and the
1908 old MTU is the lowest MTU in the path, update the route PMTU
1909 to reflect the increase. In this case if the other nodes' MTU
1910 also have the lowest MTU, TOO BIG MESSAGE will be lead to
1913 if (rt->rt6i_dev == arg->dev &&
1914 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1915 (dst_mtu(&rt->u.dst) >= arg->mtu ||
1916 (dst_mtu(&rt->u.dst) < arg->mtu &&
1917 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1918 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1919 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1924 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1926 struct rt6_mtu_change_arg arg = {
1931 fib6_clean_all(dev->nd_net, rt6_mtu_change_route, 0, &arg);
1934 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1935 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
1936 [RTA_OIF] = { .type = NLA_U32 },
1937 [RTA_IIF] = { .type = NLA_U32 },
1938 [RTA_PRIORITY] = { .type = NLA_U32 },
1939 [RTA_METRICS] = { .type = NLA_NESTED },
1942 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1943 struct fib6_config *cfg)
1946 struct nlattr *tb[RTA_MAX+1];
1949 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1954 rtm = nlmsg_data(nlh);
1955 memset(cfg, 0, sizeof(*cfg));
1957 cfg->fc_table = rtm->rtm_table;
1958 cfg->fc_dst_len = rtm->rtm_dst_len;
1959 cfg->fc_src_len = rtm->rtm_src_len;
1960 cfg->fc_flags = RTF_UP;
1961 cfg->fc_protocol = rtm->rtm_protocol;
1963 if (rtm->rtm_type == RTN_UNREACHABLE)
1964 cfg->fc_flags |= RTF_REJECT;
1966 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
1967 cfg->fc_nlinfo.nlh = nlh;
1968 cfg->fc_nlinfo.nl_net = skb->sk->sk_net;
1970 if (tb[RTA_GATEWAY]) {
1971 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1972 cfg->fc_flags |= RTF_GATEWAY;
1976 int plen = (rtm->rtm_dst_len + 7) >> 3;
1978 if (nla_len(tb[RTA_DST]) < plen)
1981 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1985 int plen = (rtm->rtm_src_len + 7) >> 3;
1987 if (nla_len(tb[RTA_SRC]) < plen)
1990 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1994 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
1996 if (tb[RTA_PRIORITY])
1997 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
1999 if (tb[RTA_METRICS]) {
2000 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2001 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2005 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2012 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2014 struct net *net = skb->sk->sk_net;
2015 struct fib6_config cfg;
2018 if (net != &init_net)
2021 err = rtm_to_fib6_config(skb, nlh, &cfg);
2025 return ip6_route_del(&cfg);
2028 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2030 struct net *net = skb->sk->sk_net;
2031 struct fib6_config cfg;
2034 if (net != &init_net)
2037 err = rtm_to_fib6_config(skb, nlh, &cfg);
2041 return ip6_route_add(&cfg);
2044 static inline size_t rt6_nlmsg_size(void)
2046 return NLMSG_ALIGN(sizeof(struct rtmsg))
2047 + nla_total_size(16) /* RTA_SRC */
2048 + nla_total_size(16) /* RTA_DST */
2049 + nla_total_size(16) /* RTA_GATEWAY */
2050 + nla_total_size(16) /* RTA_PREFSRC */
2051 + nla_total_size(4) /* RTA_TABLE */
2052 + nla_total_size(4) /* RTA_IIF */
2053 + nla_total_size(4) /* RTA_OIF */
2054 + nla_total_size(4) /* RTA_PRIORITY */
2055 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2056 + nla_total_size(sizeof(struct rta_cacheinfo));
2059 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2060 struct in6_addr *dst, struct in6_addr *src,
2061 int iif, int type, u32 pid, u32 seq,
2062 int prefix, unsigned int flags)
2065 struct nlmsghdr *nlh;
2069 if (prefix) { /* user wants prefix routes only */
2070 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2071 /* success since this is not a prefix route */
2076 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2080 rtm = nlmsg_data(nlh);
2081 rtm->rtm_family = AF_INET6;
2082 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2083 rtm->rtm_src_len = rt->rt6i_src.plen;
2086 table = rt->rt6i_table->tb6_id;
2088 table = RT6_TABLE_UNSPEC;
2089 rtm->rtm_table = table;
2090 NLA_PUT_U32(skb, RTA_TABLE, table);
2091 if (rt->rt6i_flags&RTF_REJECT)
2092 rtm->rtm_type = RTN_UNREACHABLE;
2093 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2094 rtm->rtm_type = RTN_LOCAL;
2096 rtm->rtm_type = RTN_UNICAST;
2098 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2099 rtm->rtm_protocol = rt->rt6i_protocol;
2100 if (rt->rt6i_flags&RTF_DYNAMIC)
2101 rtm->rtm_protocol = RTPROT_REDIRECT;
2102 else if (rt->rt6i_flags & RTF_ADDRCONF)
2103 rtm->rtm_protocol = RTPROT_KERNEL;
2104 else if (rt->rt6i_flags&RTF_DEFAULT)
2105 rtm->rtm_protocol = RTPROT_RA;
2107 if (rt->rt6i_flags&RTF_CACHE)
2108 rtm->rtm_flags |= RTM_F_CLONED;
2111 NLA_PUT(skb, RTA_DST, 16, dst);
2112 rtm->rtm_dst_len = 128;
2113 } else if (rtm->rtm_dst_len)
2114 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2115 #ifdef CONFIG_IPV6_SUBTREES
2117 NLA_PUT(skb, RTA_SRC, 16, src);
2118 rtm->rtm_src_len = 128;
2119 } else if (rtm->rtm_src_len)
2120 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2123 NLA_PUT_U32(skb, RTA_IIF, iif);
2125 struct in6_addr saddr_buf;
2126 if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
2127 dst, &saddr_buf) == 0)
2128 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2131 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2132 goto nla_put_failure;
2134 if (rt->u.dst.neighbour)
2135 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2138 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2140 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2142 expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2143 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2144 expires, rt->u.dst.error) < 0)
2145 goto nla_put_failure;
2147 return nlmsg_end(skb, nlh);
2150 nlmsg_cancel(skb, nlh);
2154 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2156 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2159 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2160 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2161 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2165 return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2166 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2167 prefix, NLM_F_MULTI);
2170 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2172 struct net *net = in_skb->sk->sk_net;
2173 struct nlattr *tb[RTA_MAX+1];
2174 struct rt6_info *rt;
2175 struct sk_buff *skb;
2180 if (net != &init_net)
2183 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2188 memset(&fl, 0, sizeof(fl));
2191 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2194 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2198 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2201 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2205 iif = nla_get_u32(tb[RTA_IIF]);
2208 fl.oif = nla_get_u32(tb[RTA_OIF]);
2211 struct net_device *dev;
2212 dev = __dev_get_by_index(&init_net, iif);
2219 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2225 /* Reserve room for dummy headers, this skb can pass
2226 through good chunk of routing engine.
2228 skb_reset_mac_header(skb);
2229 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2231 rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2232 skb->dst = &rt->u.dst;
2234 err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2235 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2236 nlh->nlmsg_seq, 0, 0);
2242 err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
2247 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2249 struct sk_buff *skb;
2254 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2256 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2260 err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2261 event, info->pid, seq, 0, 0);
2263 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2264 WARN_ON(err == -EMSGSIZE);
2268 err = rtnl_notify(skb, &init_net, info->pid,
2269 RTNLGRP_IPV6_ROUTE, info->nlh, gfp_any());
2272 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_ROUTE, err);
2279 #ifdef CONFIG_PROC_FS
2281 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2292 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2294 struct seq_file *m = p_arg;
2296 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2299 #ifdef CONFIG_IPV6_SUBTREES
2300 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2303 seq_puts(m, "00000000000000000000000000000000 00 ");
2306 if (rt->rt6i_nexthop) {
2307 seq_printf(m, NIP6_SEQFMT,
2308 NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2310 seq_puts(m, "00000000000000000000000000000000");
2312 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2313 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2314 rt->u.dst.__use, rt->rt6i_flags,
2315 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2319 static int ipv6_route_show(struct seq_file *m, void *v)
2321 struct net *net = (struct net *)m->private;
2322 fib6_clean_all(net, rt6_info_route, 0, m);
2326 static int ipv6_route_open(struct inode *inode, struct file *file)
2328 struct net *net = get_proc_net(inode);
2331 return single_open(file, ipv6_route_show, net);
2334 static int ipv6_route_release(struct inode *inode, struct file *file)
2336 struct seq_file *seq = file->private_data;
2337 struct net *net = seq->private;
2339 return single_release(inode, file);
2342 static const struct file_operations ipv6_route_proc_fops = {
2343 .owner = THIS_MODULE,
2344 .open = ipv6_route_open,
2346 .llseek = seq_lseek,
2347 .release = ipv6_route_release,
2350 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2352 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2353 rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2354 rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2355 rt6_stats.fib_rt_cache,
2356 atomic_read(&ip6_dst_ops.entries),
2357 rt6_stats.fib_discarded_routes);
2362 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2364 return single_open(file, rt6_stats_seq_show, NULL);
2367 static const struct file_operations rt6_stats_seq_fops = {
2368 .owner = THIS_MODULE,
2369 .open = rt6_stats_seq_open,
2371 .llseek = seq_lseek,
2372 .release = single_release,
2375 static int ipv6_route_proc_init(struct net *net)
2378 if (!proc_net_fops_create(net, "ipv6_route",
2379 0, &ipv6_route_proc_fops))
2382 if (!proc_net_fops_create(net, "rt6_stats",
2383 S_IRUGO, &rt6_stats_seq_fops))
2384 goto out_ipv6_route;
2390 proc_net_remove(net, "ipv6_route");
2394 static void ipv6_route_proc_fini(struct net *net)
2396 proc_net_remove(net, "ipv6_route");
2397 proc_net_remove(net, "rt6_stats");
2400 static inline int ipv6_route_proc_init(struct net *net)
2404 static inline void ipv6_route_proc_fini(struct net *net)
2408 #endif /* CONFIG_PROC_FS */
2410 #ifdef CONFIG_SYSCTL
2413 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2414 void __user *buffer, size_t *lenp, loff_t *ppos)
2416 int delay = init_net.ipv6.sysctl.flush_delay;
2418 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2419 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay);
2425 ctl_table ipv6_route_table_template[] = {
2427 .procname = "flush",
2428 .data = &init_net.ipv6.sysctl.flush_delay,
2429 .maxlen = sizeof(int),
2431 .proc_handler = &ipv6_sysctl_rtcache_flush
2434 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2435 .procname = "gc_thresh",
2436 .data = &ip6_dst_ops.gc_thresh,
2437 .maxlen = sizeof(int),
2439 .proc_handler = &proc_dointvec,
2442 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2443 .procname = "max_size",
2444 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2445 .maxlen = sizeof(int),
2447 .proc_handler = &proc_dointvec,
2450 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2451 .procname = "gc_min_interval",
2452 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2453 .maxlen = sizeof(int),
2455 .proc_handler = &proc_dointvec_jiffies,
2456 .strategy = &sysctl_jiffies,
2459 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2460 .procname = "gc_timeout",
2461 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2462 .maxlen = sizeof(int),
2464 .proc_handler = &proc_dointvec_jiffies,
2465 .strategy = &sysctl_jiffies,
2468 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2469 .procname = "gc_interval",
2470 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2471 .maxlen = sizeof(int),
2473 .proc_handler = &proc_dointvec_jiffies,
2474 .strategy = &sysctl_jiffies,
2477 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2478 .procname = "gc_elasticity",
2479 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2480 .maxlen = sizeof(int),
2482 .proc_handler = &proc_dointvec_jiffies,
2483 .strategy = &sysctl_jiffies,
2486 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2487 .procname = "mtu_expires",
2488 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2489 .maxlen = sizeof(int),
2491 .proc_handler = &proc_dointvec_jiffies,
2492 .strategy = &sysctl_jiffies,
2495 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2496 .procname = "min_adv_mss",
2497 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2498 .maxlen = sizeof(int),
2500 .proc_handler = &proc_dointvec_jiffies,
2501 .strategy = &sysctl_jiffies,
2504 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2505 .procname = "gc_min_interval_ms",
2506 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2507 .maxlen = sizeof(int),
2509 .proc_handler = &proc_dointvec_ms_jiffies,
2510 .strategy = &sysctl_ms_jiffies,
2515 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2517 struct ctl_table *table;
2519 table = kmemdup(ipv6_route_table_template,
2520 sizeof(ipv6_route_table_template),
2524 table[0].data = &net->ipv6.sysctl.flush_delay;
2525 /* table[1].data will be handled when we have
2526 routes per namespace */
2527 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2528 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2529 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2530 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2531 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2532 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2533 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2540 int __init ip6_route_init(void)
2544 ip6_dst_ops.kmem_cachep =
2545 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2546 SLAB_HWCACHE_ALIGN, NULL);
2547 if (!ip6_dst_ops.kmem_cachep)
2550 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
2554 goto out_kmem_cache;
2556 ret = ipv6_route_proc_init(&init_net);
2564 ret = fib6_rules_init();
2569 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2570 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2571 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2572 goto fib6_rules_init;
2579 fib6_rules_cleanup();
2583 ipv6_route_proc_fini(&init_net);
2585 rt6_ifdown(&init_net, NULL);
2588 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2592 void ip6_route_cleanup(void)
2594 fib6_rules_cleanup();
2595 ipv6_route_proc_fini(&init_net);
2597 rt6_ifdown(&init_net, NULL);
2599 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);