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(void);
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),
112 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
116 static struct dst_ops ip6_dst_blackhole_ops = {
118 .protocol = __constant_htons(ETH_P_IPV6),
119 .destroy = ip6_dst_destroy,
120 .check = ip6_dst_check,
121 .update_pmtu = ip6_rt_blackhole_update_pmtu,
122 .entry_size = sizeof(struct rt6_info),
125 struct rt6_info ip6_null_entry = {
128 .__refcnt = ATOMIC_INIT(1),
131 .error = -ENETUNREACH,
132 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
133 .input = ip6_pkt_discard,
134 .output = ip6_pkt_discard_out,
136 .path = (struct dst_entry*)&ip6_null_entry,
139 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
140 .rt6i_metric = ~(u32) 0,
141 .rt6i_ref = ATOMIC_INIT(1),
144 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
146 static int ip6_pkt_prohibit(struct sk_buff *skb);
147 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
149 struct rt6_info ip6_prohibit_entry = {
152 .__refcnt = ATOMIC_INIT(1),
156 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
157 .input = ip6_pkt_prohibit,
158 .output = ip6_pkt_prohibit_out,
160 .path = (struct dst_entry*)&ip6_prohibit_entry,
163 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
164 .rt6i_metric = ~(u32) 0,
165 .rt6i_ref = ATOMIC_INIT(1),
168 struct rt6_info ip6_blk_hole_entry = {
171 .__refcnt = ATOMIC_INIT(1),
175 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
176 .input = dst_discard,
177 .output = dst_discard,
179 .path = (struct dst_entry*)&ip6_blk_hole_entry,
182 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
183 .rt6i_metric = ~(u32) 0,
184 .rt6i_ref = ATOMIC_INIT(1),
189 /* allocate dst with ip6_dst_ops */
190 static __inline__ struct rt6_info *ip6_dst_alloc(void)
192 return (struct rt6_info *)dst_alloc(&ip6_dst_ops);
195 static void ip6_dst_destroy(struct dst_entry *dst)
197 struct rt6_info *rt = (struct rt6_info *)dst;
198 struct inet6_dev *idev = rt->rt6i_idev;
201 rt->rt6i_idev = NULL;
206 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
209 struct rt6_info *rt = (struct rt6_info *)dst;
210 struct inet6_dev *idev = rt->rt6i_idev;
211 struct net_device *loopback_dev =
212 dev->nd_net->loopback_dev;
214 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
215 struct inet6_dev *loopback_idev =
216 in6_dev_get(loopback_dev);
217 if (loopback_idev != NULL) {
218 rt->rt6i_idev = loopback_idev;
224 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
226 return (rt->rt6i_flags & RTF_EXPIRES &&
227 time_after(jiffies, rt->rt6i_expires));
230 static inline int rt6_need_strict(struct in6_addr *daddr)
232 return (ipv6_addr_type(daddr) &
233 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
237 * Route lookup. Any table->tb6_lock is implied.
240 static __inline__ struct rt6_info *rt6_device_match(struct rt6_info *rt,
244 struct rt6_info *local = NULL;
245 struct rt6_info *sprt;
248 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
249 struct net_device *dev = sprt->rt6i_dev;
250 if (dev->ifindex == oif)
252 if (dev->flags & IFF_LOOPBACK) {
253 if (sprt->rt6i_idev == NULL ||
254 sprt->rt6i_idev->dev->ifindex != oif) {
257 if (local && (!oif ||
258 local->rt6i_idev->dev->ifindex == oif))
269 return &ip6_null_entry;
274 #ifdef CONFIG_IPV6_ROUTER_PREF
275 static void rt6_probe(struct rt6_info *rt)
277 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
279 * Okay, this does not seem to be appropriate
280 * for now, however, we need to check if it
281 * is really so; aka Router Reachability Probing.
283 * Router Reachability Probe MUST be rate-limited
284 * to no more than one per minute.
286 if (!neigh || (neigh->nud_state & NUD_VALID))
288 read_lock_bh(&neigh->lock);
289 if (!(neigh->nud_state & NUD_VALID) &&
290 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
291 struct in6_addr mcaddr;
292 struct in6_addr *target;
294 neigh->updated = jiffies;
295 read_unlock_bh(&neigh->lock);
297 target = (struct in6_addr *)&neigh->primary_key;
298 addrconf_addr_solict_mult(target, &mcaddr);
299 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
301 read_unlock_bh(&neigh->lock);
304 static inline void rt6_probe(struct rt6_info *rt)
311 * Default Router Selection (RFC 2461 6.3.6)
313 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
315 struct net_device *dev = rt->rt6i_dev;
316 if (!oif || dev->ifindex == oif)
318 if ((dev->flags & IFF_LOOPBACK) &&
319 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
324 static inline int rt6_check_neigh(struct rt6_info *rt)
326 struct neighbour *neigh = rt->rt6i_nexthop;
328 if (rt->rt6i_flags & RTF_NONEXTHOP ||
329 !(rt->rt6i_flags & RTF_GATEWAY))
332 read_lock_bh(&neigh->lock);
333 if (neigh->nud_state & NUD_VALID)
335 #ifdef CONFIG_IPV6_ROUTER_PREF
336 else if (neigh->nud_state & NUD_FAILED)
341 read_unlock_bh(&neigh->lock);
347 static int rt6_score_route(struct rt6_info *rt, int oif,
352 m = rt6_check_dev(rt, oif);
353 if (!m && (strict & RT6_LOOKUP_F_IFACE))
355 #ifdef CONFIG_IPV6_ROUTER_PREF
356 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
358 n = rt6_check_neigh(rt);
359 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
364 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
365 int *mpri, struct rt6_info *match)
369 if (rt6_check_expired(rt))
372 m = rt6_score_route(rt, oif, strict);
377 if (strict & RT6_LOOKUP_F_REACHABLE)
381 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
389 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
390 struct rt6_info *rr_head,
391 u32 metric, int oif, int strict)
393 struct rt6_info *rt, *match;
397 for (rt = rr_head; rt && rt->rt6i_metric == metric;
398 rt = rt->u.dst.rt6_next)
399 match = find_match(rt, oif, strict, &mpri, match);
400 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
401 rt = rt->u.dst.rt6_next)
402 match = find_match(rt, oif, strict, &mpri, match);
407 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
409 struct rt6_info *match, *rt0;
411 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
412 __FUNCTION__, fn->leaf, oif);
416 fn->rr_ptr = rt0 = fn->leaf;
418 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
421 (strict & RT6_LOOKUP_F_REACHABLE)) {
422 struct rt6_info *next = rt0->u.dst.rt6_next;
424 /* no entries matched; do round-robin */
425 if (!next || next->rt6i_metric != rt0->rt6i_metric)
432 RT6_TRACE("%s() => %p\n",
433 __FUNCTION__, match);
435 return (match ? match : &ip6_null_entry);
438 #ifdef CONFIG_IPV6_ROUTE_INFO
439 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
440 struct in6_addr *gwaddr)
442 struct route_info *rinfo = (struct route_info *) opt;
443 struct in6_addr prefix_buf, *prefix;
448 if (len < sizeof(struct route_info)) {
452 /* Sanity check for prefix_len and length */
453 if (rinfo->length > 3) {
455 } else if (rinfo->prefix_len > 128) {
457 } else if (rinfo->prefix_len > 64) {
458 if (rinfo->length < 2) {
461 } else if (rinfo->prefix_len > 0) {
462 if (rinfo->length < 1) {
467 pref = rinfo->route_pref;
468 if (pref == ICMPV6_ROUTER_PREF_INVALID)
469 pref = ICMPV6_ROUTER_PREF_MEDIUM;
471 lifetime = ntohl(rinfo->lifetime);
472 if (lifetime == 0xffffffff) {
474 } else if (lifetime > 0x7fffffff/HZ) {
475 /* Avoid arithmetic overflow */
476 lifetime = 0x7fffffff/HZ - 1;
479 if (rinfo->length == 3)
480 prefix = (struct in6_addr *)rinfo->prefix;
482 /* this function is safe */
483 ipv6_addr_prefix(&prefix_buf,
484 (struct in6_addr *)rinfo->prefix,
486 prefix = &prefix_buf;
489 rt = rt6_get_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex);
491 if (rt && !lifetime) {
497 rt = rt6_add_route_info(prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
500 rt->rt6i_flags = RTF_ROUTEINFO |
501 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
504 if (lifetime == 0xffffffff) {
505 rt->rt6i_flags &= ~RTF_EXPIRES;
507 rt->rt6i_expires = jiffies + HZ * lifetime;
508 rt->rt6i_flags |= RTF_EXPIRES;
510 dst_release(&rt->u.dst);
516 #define BACKTRACK(saddr) \
518 if (rt == &ip6_null_entry) { \
519 struct fib6_node *pn; \
521 if (fn->fn_flags & RTN_TL_ROOT) \
524 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
525 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
528 if (fn->fn_flags & RTN_RTINFO) \
534 static struct rt6_info *ip6_pol_route_lookup(struct fib6_table *table,
535 struct flowi *fl, int flags)
537 struct fib6_node *fn;
540 read_lock_bh(&table->tb6_lock);
541 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
544 rt = rt6_device_match(rt, fl->oif, flags);
545 BACKTRACK(&fl->fl6_src);
547 dst_use(&rt->u.dst, jiffies);
548 read_unlock_bh(&table->tb6_lock);
553 struct rt6_info *rt6_lookup(struct in6_addr *daddr, struct in6_addr *saddr,
564 struct dst_entry *dst;
565 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
568 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
569 flags |= RT6_LOOKUP_F_HAS_SADDR;
572 dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_lookup);
574 return (struct rt6_info *) dst;
581 EXPORT_SYMBOL(rt6_lookup);
583 /* ip6_ins_rt is called with FREE table->tb6_lock.
584 It takes new route entry, the addition fails by any reason the
585 route is freed. In any case, if caller does not hold it, it may
589 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
592 struct fib6_table *table;
594 table = rt->rt6i_table;
595 write_lock_bh(&table->tb6_lock);
596 err = fib6_add(&table->tb6_root, rt, info);
597 write_unlock_bh(&table->tb6_lock);
602 int ip6_ins_rt(struct rt6_info *rt)
604 struct nl_info info = {};
605 return __ip6_ins_rt(rt, &info);
608 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
609 struct in6_addr *saddr)
617 rt = ip6_rt_copy(ort);
620 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
621 if (rt->rt6i_dst.plen != 128 &&
622 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
623 rt->rt6i_flags |= RTF_ANYCAST;
624 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
627 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
628 rt->rt6i_dst.plen = 128;
629 rt->rt6i_flags |= RTF_CACHE;
630 rt->u.dst.flags |= DST_HOST;
632 #ifdef CONFIG_IPV6_SUBTREES
633 if (rt->rt6i_src.plen && saddr) {
634 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
635 rt->rt6i_src.plen = 128;
639 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
646 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
648 struct rt6_info *rt = ip6_rt_copy(ort);
650 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
651 rt->rt6i_dst.plen = 128;
652 rt->rt6i_flags |= RTF_CACHE;
653 rt->u.dst.flags |= DST_HOST;
654 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
659 static struct rt6_info *ip6_pol_route(struct fib6_table *table, int oif,
660 struct flowi *fl, int flags)
662 struct fib6_node *fn;
663 struct rt6_info *rt, *nrt;
667 int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
669 strict |= flags & RT6_LOOKUP_F_IFACE;
672 read_lock_bh(&table->tb6_lock);
675 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
678 rt = rt6_select(fn, oif, strict | reachable);
679 BACKTRACK(&fl->fl6_src);
680 if (rt == &ip6_null_entry ||
681 rt->rt6i_flags & RTF_CACHE)
684 dst_hold(&rt->u.dst);
685 read_unlock_bh(&table->tb6_lock);
687 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
688 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
690 #if CLONE_OFFLINK_ROUTE
691 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
697 dst_release(&rt->u.dst);
698 rt = nrt ? : &ip6_null_entry;
700 dst_hold(&rt->u.dst);
702 err = ip6_ins_rt(nrt);
711 * Race condition! In the gap, when table->tb6_lock was
712 * released someone could insert this route. Relookup.
714 dst_release(&rt->u.dst);
722 dst_hold(&rt->u.dst);
723 read_unlock_bh(&table->tb6_lock);
725 rt->u.dst.lastuse = jiffies;
731 static struct rt6_info *ip6_pol_route_input(struct fib6_table *table,
732 struct flowi *fl, int flags)
734 return ip6_pol_route(table, fl->iif, fl, flags);
737 void ip6_route_input(struct sk_buff *skb)
739 struct ipv6hdr *iph = ipv6_hdr(skb);
740 int flags = RT6_LOOKUP_F_HAS_SADDR;
742 .iif = skb->dev->ifindex,
747 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
751 .proto = iph->nexthdr,
754 if (rt6_need_strict(&iph->daddr))
755 flags |= RT6_LOOKUP_F_IFACE;
757 skb->dst = fib6_rule_lookup(&fl, flags, ip6_pol_route_input);
760 static struct rt6_info *ip6_pol_route_output(struct fib6_table *table,
761 struct flowi *fl, int flags)
763 return ip6_pol_route(table, fl->oif, fl, flags);
766 struct dst_entry * ip6_route_output(struct sock *sk, struct flowi *fl)
770 if (rt6_need_strict(&fl->fl6_dst))
771 flags |= RT6_LOOKUP_F_IFACE;
773 if (!ipv6_addr_any(&fl->fl6_src))
774 flags |= RT6_LOOKUP_F_HAS_SADDR;
776 return fib6_rule_lookup(fl, flags, ip6_pol_route_output);
779 EXPORT_SYMBOL(ip6_route_output);
781 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
783 struct rt6_info *ort = (struct rt6_info *) *dstp;
784 struct rt6_info *rt = (struct rt6_info *)
785 dst_alloc(&ip6_dst_blackhole_ops);
786 struct dst_entry *new = NULL;
791 atomic_set(&new->__refcnt, 1);
793 new->input = dst_discard;
794 new->output = dst_discard;
796 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
797 new->dev = ort->u.dst.dev;
800 rt->rt6i_idev = ort->rt6i_idev;
802 in6_dev_hold(rt->rt6i_idev);
803 rt->rt6i_expires = 0;
805 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
806 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
809 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
810 #ifdef CONFIG_IPV6_SUBTREES
811 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
819 return (new ? 0 : -ENOMEM);
821 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
824 * Destination cache support functions
827 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
831 rt = (struct rt6_info *) dst;
833 if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
839 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
841 struct rt6_info *rt = (struct rt6_info *) dst;
844 if (rt->rt6i_flags & RTF_CACHE)
852 static void ip6_link_failure(struct sk_buff *skb)
856 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
858 rt = (struct rt6_info *) skb->dst;
860 if (rt->rt6i_flags&RTF_CACHE) {
861 dst_set_expires(&rt->u.dst, 0);
862 rt->rt6i_flags |= RTF_EXPIRES;
863 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
864 rt->rt6i_node->fn_sernum = -1;
868 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
870 struct rt6_info *rt6 = (struct rt6_info*)dst;
872 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
873 rt6->rt6i_flags |= RTF_MODIFIED;
874 if (mtu < IPV6_MIN_MTU) {
876 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
878 dst->metrics[RTAX_MTU-1] = mtu;
879 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
883 static int ipv6_get_mtu(struct net_device *dev);
885 static inline unsigned int ipv6_advmss(unsigned int mtu)
887 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
889 if (mtu < init_net.ipv6.sysctl.ip6_rt_min_advmss)
890 mtu = init_net.ipv6.sysctl.ip6_rt_min_advmss;
893 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
894 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
895 * IPV6_MAXPLEN is also valid and means: "any MSS,
896 * rely only on pmtu discovery"
898 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
903 static struct dst_entry *ndisc_dst_gc_list;
904 static DEFINE_SPINLOCK(ndisc_lock);
906 struct dst_entry *ndisc_dst_alloc(struct net_device *dev,
907 struct neighbour *neigh,
908 struct in6_addr *addr,
909 int (*output)(struct sk_buff *))
912 struct inet6_dev *idev = in6_dev_get(dev);
914 if (unlikely(idev == NULL))
917 rt = ip6_dst_alloc();
918 if (unlikely(rt == NULL)) {
927 neigh = ndisc_get_neigh(dev, addr);
930 rt->rt6i_idev = idev;
931 rt->rt6i_nexthop = neigh;
932 atomic_set(&rt->u.dst.__refcnt, 1);
933 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
934 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
935 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
936 rt->u.dst.output = output;
938 #if 0 /* there's no chance to use these for ndisc */
939 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
942 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
943 rt->rt6i_dst.plen = 128;
946 spin_lock_bh(&ndisc_lock);
947 rt->u.dst.next = ndisc_dst_gc_list;
948 ndisc_dst_gc_list = &rt->u.dst;
949 spin_unlock_bh(&ndisc_lock);
951 fib6_force_start_gc();
957 int ndisc_dst_gc(int *more)
959 struct dst_entry *dst, *next, **pprev;
965 spin_lock_bh(&ndisc_lock);
966 pprev = &ndisc_dst_gc_list;
968 while ((dst = *pprev) != NULL) {
969 if (!atomic_read(&dst->__refcnt)) {
979 spin_unlock_bh(&ndisc_lock);
984 static int ip6_dst_gc(void)
986 static unsigned expire = 30*HZ;
987 static unsigned long last_gc;
988 unsigned long now = jiffies;
990 if (time_after(last_gc + init_net.ipv6.sysctl.ip6_rt_gc_min_interval, now) &&
991 atomic_read(&ip6_dst_ops.entries) <= init_net.ipv6.sysctl.ip6_rt_max_size)
997 if (atomic_read(&ip6_dst_ops.entries) < ip6_dst_ops.gc_thresh)
998 expire = init_net.ipv6.sysctl.ip6_rt_gc_timeout>>1;
1001 expire -= expire>>init_net.ipv6.sysctl.ip6_rt_gc_elasticity;
1002 return (atomic_read(&ip6_dst_ops.entries) > init_net.ipv6.sysctl.ip6_rt_max_size);
1005 /* Clean host part of a prefix. Not necessary in radix tree,
1006 but results in cleaner routing tables.
1008 Remove it only when all the things will work!
1011 static int ipv6_get_mtu(struct net_device *dev)
1013 int mtu = IPV6_MIN_MTU;
1014 struct inet6_dev *idev;
1016 idev = in6_dev_get(dev);
1018 mtu = idev->cnf.mtu6;
1024 int ipv6_get_hoplimit(struct net_device *dev)
1026 int hoplimit = ipv6_devconf.hop_limit;
1027 struct inet6_dev *idev;
1029 idev = in6_dev_get(dev);
1031 hoplimit = idev->cnf.hop_limit;
1041 int ip6_route_add(struct fib6_config *cfg)
1044 struct rt6_info *rt = NULL;
1045 struct net_device *dev = NULL;
1046 struct inet6_dev *idev = NULL;
1047 struct fib6_table *table;
1050 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1052 #ifndef CONFIG_IPV6_SUBTREES
1053 if (cfg->fc_src_len)
1056 if (cfg->fc_ifindex) {
1058 dev = dev_get_by_index(&init_net, cfg->fc_ifindex);
1061 idev = in6_dev_get(dev);
1066 if (cfg->fc_metric == 0)
1067 cfg->fc_metric = IP6_RT_PRIO_USER;
1069 table = fib6_new_table(cfg->fc_table);
1070 if (table == NULL) {
1075 rt = ip6_dst_alloc();
1082 rt->u.dst.obsolete = -1;
1083 rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1085 if (cfg->fc_protocol == RTPROT_UNSPEC)
1086 cfg->fc_protocol = RTPROT_BOOT;
1087 rt->rt6i_protocol = cfg->fc_protocol;
1089 addr_type = ipv6_addr_type(&cfg->fc_dst);
1091 if (addr_type & IPV6_ADDR_MULTICAST)
1092 rt->u.dst.input = ip6_mc_input;
1094 rt->u.dst.input = ip6_forward;
1096 rt->u.dst.output = ip6_output;
1098 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1099 rt->rt6i_dst.plen = cfg->fc_dst_len;
1100 if (rt->rt6i_dst.plen == 128)
1101 rt->u.dst.flags = DST_HOST;
1103 #ifdef CONFIG_IPV6_SUBTREES
1104 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1105 rt->rt6i_src.plen = cfg->fc_src_len;
1108 rt->rt6i_metric = cfg->fc_metric;
1110 /* We cannot add true routes via loopback here,
1111 they would result in kernel looping; promote them to reject routes
1113 if ((cfg->fc_flags & RTF_REJECT) ||
1114 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1115 /* hold loopback dev/idev if we haven't done so. */
1116 if (dev != init_net.loopback_dev) {
1121 dev = init_net.loopback_dev;
1123 idev = in6_dev_get(dev);
1129 rt->u.dst.output = ip6_pkt_discard_out;
1130 rt->u.dst.input = ip6_pkt_discard;
1131 rt->u.dst.error = -ENETUNREACH;
1132 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1136 if (cfg->fc_flags & RTF_GATEWAY) {
1137 struct in6_addr *gw_addr;
1140 gw_addr = &cfg->fc_gateway;
1141 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1142 gwa_type = ipv6_addr_type(gw_addr);
1144 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1145 struct rt6_info *grt;
1147 /* IPv6 strictly inhibits using not link-local
1148 addresses as nexthop address.
1149 Otherwise, router will not able to send redirects.
1150 It is very good, but in some (rare!) circumstances
1151 (SIT, PtP, NBMA NOARP links) it is handy to allow
1152 some exceptions. --ANK
1155 if (!(gwa_type&IPV6_ADDR_UNICAST))
1158 grt = rt6_lookup(gw_addr, NULL, cfg->fc_ifindex, 1);
1160 err = -EHOSTUNREACH;
1164 if (dev != grt->rt6i_dev) {
1165 dst_release(&grt->u.dst);
1169 dev = grt->rt6i_dev;
1170 idev = grt->rt6i_idev;
1172 in6_dev_hold(grt->rt6i_idev);
1174 if (!(grt->rt6i_flags&RTF_GATEWAY))
1176 dst_release(&grt->u.dst);
1182 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1190 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1191 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1192 if (IS_ERR(rt->rt6i_nexthop)) {
1193 err = PTR_ERR(rt->rt6i_nexthop);
1194 rt->rt6i_nexthop = NULL;
1199 rt->rt6i_flags = cfg->fc_flags;
1206 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1207 int type = nla_type(nla);
1210 if (type > RTAX_MAX) {
1215 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1220 if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1221 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1222 if (!rt->u.dst.metrics[RTAX_MTU-1])
1223 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1224 if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1225 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1226 rt->u.dst.dev = dev;
1227 rt->rt6i_idev = idev;
1228 rt->rt6i_table = table;
1229 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1237 dst_free(&rt->u.dst);
1241 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1244 struct fib6_table *table;
1246 if (rt == &ip6_null_entry)
1249 table = rt->rt6i_table;
1250 write_lock_bh(&table->tb6_lock);
1252 err = fib6_del(rt, info);
1253 dst_release(&rt->u.dst);
1255 write_unlock_bh(&table->tb6_lock);
1260 int ip6_del_rt(struct rt6_info *rt)
1262 struct nl_info info = {};
1263 return __ip6_del_rt(rt, &info);
1266 static int ip6_route_del(struct fib6_config *cfg)
1268 struct fib6_table *table;
1269 struct fib6_node *fn;
1270 struct rt6_info *rt;
1273 table = fib6_get_table(cfg->fc_table);
1277 read_lock_bh(&table->tb6_lock);
1279 fn = fib6_locate(&table->tb6_root,
1280 &cfg->fc_dst, cfg->fc_dst_len,
1281 &cfg->fc_src, cfg->fc_src_len);
1284 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1285 if (cfg->fc_ifindex &&
1286 (rt->rt6i_dev == NULL ||
1287 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1289 if (cfg->fc_flags & RTF_GATEWAY &&
1290 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1292 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1294 dst_hold(&rt->u.dst);
1295 read_unlock_bh(&table->tb6_lock);
1297 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1300 read_unlock_bh(&table->tb6_lock);
1308 struct ip6rd_flowi {
1310 struct in6_addr gateway;
1313 static struct rt6_info *__ip6_route_redirect(struct fib6_table *table,
1317 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1318 struct rt6_info *rt;
1319 struct fib6_node *fn;
1322 * Get the "current" route for this destination and
1323 * check if the redirect has come from approriate router.
1325 * RFC 2461 specifies that redirects should only be
1326 * accepted if they come from the nexthop to the target.
1327 * Due to the way the routes are chosen, this notion
1328 * is a bit fuzzy and one might need to check all possible
1332 read_lock_bh(&table->tb6_lock);
1333 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1335 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1337 * Current route is on-link; redirect is always invalid.
1339 * Seems, previous statement is not true. It could
1340 * be node, which looks for us as on-link (f.e. proxy ndisc)
1341 * But then router serving it might decide, that we should
1342 * know truth 8)8) --ANK (980726).
1344 if (rt6_check_expired(rt))
1346 if (!(rt->rt6i_flags & RTF_GATEWAY))
1348 if (fl->oif != rt->rt6i_dev->ifindex)
1350 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1356 rt = &ip6_null_entry;
1357 BACKTRACK(&fl->fl6_src);
1359 dst_hold(&rt->u.dst);
1361 read_unlock_bh(&table->tb6_lock);
1366 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1367 struct in6_addr *src,
1368 struct in6_addr *gateway,
1369 struct net_device *dev)
1371 int flags = RT6_LOOKUP_F_HAS_SADDR;
1372 struct ip6rd_flowi rdfl = {
1374 .oif = dev->ifindex,
1382 .gateway = *gateway,
1385 if (rt6_need_strict(dest))
1386 flags |= RT6_LOOKUP_F_IFACE;
1388 return (struct rt6_info *)fib6_rule_lookup((struct flowi *)&rdfl, flags, __ip6_route_redirect);
1391 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1392 struct in6_addr *saddr,
1393 struct neighbour *neigh, u8 *lladdr, int on_link)
1395 struct rt6_info *rt, *nrt = NULL;
1396 struct netevent_redirect netevent;
1398 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1400 if (rt == &ip6_null_entry) {
1401 if (net_ratelimit())
1402 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1403 "for redirect target\n");
1408 * We have finally decided to accept it.
1411 neigh_update(neigh, lladdr, NUD_STALE,
1412 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1413 NEIGH_UPDATE_F_OVERRIDE|
1414 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1415 NEIGH_UPDATE_F_ISROUTER))
1419 * Redirect received -> path was valid.
1420 * Look, redirects are sent only in response to data packets,
1421 * so that this nexthop apparently is reachable. --ANK
1423 dst_confirm(&rt->u.dst);
1425 /* Duplicate redirect: silently ignore. */
1426 if (neigh == rt->u.dst.neighbour)
1429 nrt = ip6_rt_copy(rt);
1433 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1435 nrt->rt6i_flags &= ~RTF_GATEWAY;
1437 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1438 nrt->rt6i_dst.plen = 128;
1439 nrt->u.dst.flags |= DST_HOST;
1441 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1442 nrt->rt6i_nexthop = neigh_clone(neigh);
1443 /* Reset pmtu, it may be better */
1444 nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1445 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&nrt->u.dst));
1447 if (ip6_ins_rt(nrt))
1450 netevent.old = &rt->u.dst;
1451 netevent.new = &nrt->u.dst;
1452 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1454 if (rt->rt6i_flags&RTF_CACHE) {
1460 dst_release(&rt->u.dst);
1465 * Handle ICMP "packet too big" messages
1466 * i.e. Path MTU discovery
1469 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1470 struct net_device *dev, u32 pmtu)
1472 struct rt6_info *rt, *nrt;
1475 rt = rt6_lookup(daddr, saddr, dev->ifindex, 0);
1479 if (pmtu >= dst_mtu(&rt->u.dst))
1482 if (pmtu < IPV6_MIN_MTU) {
1484 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1485 * MTU (1280) and a fragment header should always be included
1486 * after a node receiving Too Big message reporting PMTU is
1487 * less than the IPv6 Minimum Link MTU.
1489 pmtu = IPV6_MIN_MTU;
1493 /* New mtu received -> path was valid.
1494 They are sent only in response to data packets,
1495 so that this nexthop apparently is reachable. --ANK
1497 dst_confirm(&rt->u.dst);
1499 /* Host route. If it is static, it would be better
1500 not to override it, but add new one, so that
1501 when cache entry will expire old pmtu
1502 would return automatically.
1504 if (rt->rt6i_flags & RTF_CACHE) {
1505 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1507 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1508 dst_set_expires(&rt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1509 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1514 Two cases are possible:
1515 1. It is connected route. Action: COW
1516 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1518 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1519 nrt = rt6_alloc_cow(rt, daddr, saddr);
1521 nrt = rt6_alloc_clone(rt, daddr);
1524 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1526 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1528 /* According to RFC 1981, detecting PMTU increase shouldn't be
1529 * happened within 5 mins, the recommended timer is 10 mins.
1530 * Here this route expiration time is set to ip6_rt_mtu_expires
1531 * which is 10 mins. After 10 mins the decreased pmtu is expired
1532 * and detecting PMTU increase will be automatically happened.
1534 dst_set_expires(&nrt->u.dst, init_net.ipv6.sysctl.ip6_rt_mtu_expires);
1535 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1540 dst_release(&rt->u.dst);
1544 * Misc support functions
1547 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1549 struct rt6_info *rt = ip6_dst_alloc();
1552 rt->u.dst.input = ort->u.dst.input;
1553 rt->u.dst.output = ort->u.dst.output;
1555 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1556 rt->u.dst.error = ort->u.dst.error;
1557 rt->u.dst.dev = ort->u.dst.dev;
1559 dev_hold(rt->u.dst.dev);
1560 rt->rt6i_idev = ort->rt6i_idev;
1562 in6_dev_hold(rt->rt6i_idev);
1563 rt->u.dst.lastuse = jiffies;
1564 rt->rt6i_expires = 0;
1566 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1567 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1568 rt->rt6i_metric = 0;
1570 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1571 #ifdef CONFIG_IPV6_SUBTREES
1572 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1574 rt->rt6i_table = ort->rt6i_table;
1579 #ifdef CONFIG_IPV6_ROUTE_INFO
1580 static struct rt6_info *rt6_get_route_info(struct in6_addr *prefix, int prefixlen,
1581 struct in6_addr *gwaddr, int ifindex)
1583 struct fib6_node *fn;
1584 struct rt6_info *rt = NULL;
1585 struct fib6_table *table;
1587 table = fib6_get_table(RT6_TABLE_INFO);
1591 write_lock_bh(&table->tb6_lock);
1592 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1596 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1597 if (rt->rt6i_dev->ifindex != ifindex)
1599 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1601 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1603 dst_hold(&rt->u.dst);
1607 write_unlock_bh(&table->tb6_lock);
1611 static struct rt6_info *rt6_add_route_info(struct in6_addr *prefix, int prefixlen,
1612 struct in6_addr *gwaddr, int ifindex,
1615 struct fib6_config cfg = {
1616 .fc_table = RT6_TABLE_INFO,
1618 .fc_ifindex = ifindex,
1619 .fc_dst_len = prefixlen,
1620 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1621 RTF_UP | RTF_PREF(pref),
1624 ipv6_addr_copy(&cfg.fc_dst, prefix);
1625 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1627 /* We should treat it as a default route if prefix length is 0. */
1629 cfg.fc_flags |= RTF_DEFAULT;
1631 ip6_route_add(&cfg);
1633 return rt6_get_route_info(prefix, prefixlen, gwaddr, ifindex);
1637 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1639 struct rt6_info *rt;
1640 struct fib6_table *table;
1642 table = fib6_get_table(RT6_TABLE_DFLT);
1646 write_lock_bh(&table->tb6_lock);
1647 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1648 if (dev == rt->rt6i_dev &&
1649 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1650 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1654 dst_hold(&rt->u.dst);
1655 write_unlock_bh(&table->tb6_lock);
1659 EXPORT_SYMBOL(rt6_get_dflt_router);
1661 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1662 struct net_device *dev,
1665 struct fib6_config cfg = {
1666 .fc_table = RT6_TABLE_DFLT,
1668 .fc_ifindex = dev->ifindex,
1669 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1670 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1673 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1675 ip6_route_add(&cfg);
1677 return rt6_get_dflt_router(gwaddr, dev);
1680 void rt6_purge_dflt_routers(void)
1682 struct rt6_info *rt;
1683 struct fib6_table *table;
1685 /* NOTE: Keep consistent with rt6_get_dflt_router */
1686 table = fib6_get_table(RT6_TABLE_DFLT);
1691 read_lock_bh(&table->tb6_lock);
1692 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1693 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1694 dst_hold(&rt->u.dst);
1695 read_unlock_bh(&table->tb6_lock);
1700 read_unlock_bh(&table->tb6_lock);
1703 static void rtmsg_to_fib6_config(struct in6_rtmsg *rtmsg,
1704 struct fib6_config *cfg)
1706 memset(cfg, 0, sizeof(*cfg));
1708 cfg->fc_table = RT6_TABLE_MAIN;
1709 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1710 cfg->fc_metric = rtmsg->rtmsg_metric;
1711 cfg->fc_expires = rtmsg->rtmsg_info;
1712 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1713 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1714 cfg->fc_flags = rtmsg->rtmsg_flags;
1716 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1717 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1718 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1721 int ipv6_route_ioctl(unsigned int cmd, void __user *arg)
1723 struct fib6_config cfg;
1724 struct in6_rtmsg rtmsg;
1728 case SIOCADDRT: /* Add a route */
1729 case SIOCDELRT: /* Delete a route */
1730 if (!capable(CAP_NET_ADMIN))
1732 err = copy_from_user(&rtmsg, arg,
1733 sizeof(struct in6_rtmsg));
1737 rtmsg_to_fib6_config(&rtmsg, &cfg);
1742 err = ip6_route_add(&cfg);
1745 err = ip6_route_del(&cfg);
1759 * Drop the packet on the floor
1762 static inline int ip6_pkt_drop(struct sk_buff *skb, int code,
1763 int ipstats_mib_noroutes)
1766 switch (ipstats_mib_noroutes) {
1767 case IPSTATS_MIB_INNOROUTES:
1768 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1769 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1770 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1774 case IPSTATS_MIB_OUTNOROUTES:
1775 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1778 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1783 static int ip6_pkt_discard(struct sk_buff *skb)
1785 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1788 static int ip6_pkt_discard_out(struct sk_buff *skb)
1790 skb->dev = skb->dst->dev;
1791 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1794 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1796 static int ip6_pkt_prohibit(struct sk_buff *skb)
1798 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1801 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1803 skb->dev = skb->dst->dev;
1804 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1810 * Allocate a dst for local (unicast / anycast) address.
1813 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1814 const struct in6_addr *addr,
1817 struct rt6_info *rt = ip6_dst_alloc();
1820 return ERR_PTR(-ENOMEM);
1822 dev_hold(init_net.loopback_dev);
1825 rt->u.dst.flags = DST_HOST;
1826 rt->u.dst.input = ip6_input;
1827 rt->u.dst.output = ip6_output;
1828 rt->rt6i_dev = init_net.loopback_dev;
1829 rt->rt6i_idev = idev;
1830 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1831 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dst_mtu(&rt->u.dst));
1832 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1833 rt->u.dst.obsolete = -1;
1835 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1837 rt->rt6i_flags |= RTF_ANYCAST;
1839 rt->rt6i_flags |= RTF_LOCAL;
1840 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1841 if (rt->rt6i_nexthop == NULL) {
1842 dst_free(&rt->u.dst);
1843 return ERR_PTR(-ENOMEM);
1846 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1847 rt->rt6i_dst.plen = 128;
1848 rt->rt6i_table = fib6_get_table(RT6_TABLE_LOCAL);
1850 atomic_set(&rt->u.dst.__refcnt, 1);
1855 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1857 if (((void*)rt->rt6i_dev == arg || arg == NULL) &&
1858 rt != &ip6_null_entry) {
1859 RT6_TRACE("deleted by ifdown %p\n", rt);
1865 void rt6_ifdown(struct net_device *dev)
1867 fib6_clean_all(fib6_ifdown, 0, dev);
1870 struct rt6_mtu_change_arg
1872 struct net_device *dev;
1876 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1878 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1879 struct inet6_dev *idev;
1881 /* In IPv6 pmtu discovery is not optional,
1882 so that RTAX_MTU lock cannot disable it.
1883 We still use this lock to block changes
1884 caused by addrconf/ndisc.
1887 idev = __in6_dev_get(arg->dev);
1891 /* For administrative MTU increase, there is no way to discover
1892 IPv6 PMTU increase, so PMTU increase should be updated here.
1893 Since RFC 1981 doesn't include administrative MTU increase
1894 update PMTU increase is a MUST. (i.e. jumbo frame)
1897 If new MTU is less than route PMTU, this new MTU will be the
1898 lowest MTU in the path, update the route PMTU to reflect PMTU
1899 decreases; if new MTU is greater than route PMTU, and the
1900 old MTU is the lowest MTU in the path, update the route PMTU
1901 to reflect the increase. In this case if the other nodes' MTU
1902 also have the lowest MTU, TOO BIG MESSAGE will be lead to
1905 if (rt->rt6i_dev == arg->dev &&
1906 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1907 (dst_mtu(&rt->u.dst) > arg->mtu ||
1908 (dst_mtu(&rt->u.dst) < arg->mtu &&
1909 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1910 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1911 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(arg->mtu);
1916 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1918 struct rt6_mtu_change_arg arg = {
1923 fib6_clean_all(rt6_mtu_change_route, 0, &arg);
1926 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1927 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
1928 [RTA_OIF] = { .type = NLA_U32 },
1929 [RTA_IIF] = { .type = NLA_U32 },
1930 [RTA_PRIORITY] = { .type = NLA_U32 },
1931 [RTA_METRICS] = { .type = NLA_NESTED },
1934 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1935 struct fib6_config *cfg)
1938 struct nlattr *tb[RTA_MAX+1];
1941 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
1946 rtm = nlmsg_data(nlh);
1947 memset(cfg, 0, sizeof(*cfg));
1949 cfg->fc_table = rtm->rtm_table;
1950 cfg->fc_dst_len = rtm->rtm_dst_len;
1951 cfg->fc_src_len = rtm->rtm_src_len;
1952 cfg->fc_flags = RTF_UP;
1953 cfg->fc_protocol = rtm->rtm_protocol;
1955 if (rtm->rtm_type == RTN_UNREACHABLE)
1956 cfg->fc_flags |= RTF_REJECT;
1958 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
1959 cfg->fc_nlinfo.nlh = nlh;
1961 if (tb[RTA_GATEWAY]) {
1962 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
1963 cfg->fc_flags |= RTF_GATEWAY;
1967 int plen = (rtm->rtm_dst_len + 7) >> 3;
1969 if (nla_len(tb[RTA_DST]) < plen)
1972 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
1976 int plen = (rtm->rtm_src_len + 7) >> 3;
1978 if (nla_len(tb[RTA_SRC]) < plen)
1981 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
1985 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
1987 if (tb[RTA_PRIORITY])
1988 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
1990 if (tb[RTA_METRICS]) {
1991 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
1992 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
1996 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2003 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2005 struct net *net = skb->sk->sk_net;
2006 struct fib6_config cfg;
2009 if (net != &init_net)
2012 err = rtm_to_fib6_config(skb, nlh, &cfg);
2016 return ip6_route_del(&cfg);
2019 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2021 struct net *net = skb->sk->sk_net;
2022 struct fib6_config cfg;
2025 if (net != &init_net)
2028 err = rtm_to_fib6_config(skb, nlh, &cfg);
2032 return ip6_route_add(&cfg);
2035 static inline size_t rt6_nlmsg_size(void)
2037 return NLMSG_ALIGN(sizeof(struct rtmsg))
2038 + nla_total_size(16) /* RTA_SRC */
2039 + nla_total_size(16) /* RTA_DST */
2040 + nla_total_size(16) /* RTA_GATEWAY */
2041 + nla_total_size(16) /* RTA_PREFSRC */
2042 + nla_total_size(4) /* RTA_TABLE */
2043 + nla_total_size(4) /* RTA_IIF */
2044 + nla_total_size(4) /* RTA_OIF */
2045 + nla_total_size(4) /* RTA_PRIORITY */
2046 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2047 + nla_total_size(sizeof(struct rta_cacheinfo));
2050 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2051 struct in6_addr *dst, struct in6_addr *src,
2052 int iif, int type, u32 pid, u32 seq,
2053 int prefix, unsigned int flags)
2056 struct nlmsghdr *nlh;
2060 if (prefix) { /* user wants prefix routes only */
2061 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2062 /* success since this is not a prefix route */
2067 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2071 rtm = nlmsg_data(nlh);
2072 rtm->rtm_family = AF_INET6;
2073 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2074 rtm->rtm_src_len = rt->rt6i_src.plen;
2077 table = rt->rt6i_table->tb6_id;
2079 table = RT6_TABLE_UNSPEC;
2080 rtm->rtm_table = table;
2081 NLA_PUT_U32(skb, RTA_TABLE, table);
2082 if (rt->rt6i_flags&RTF_REJECT)
2083 rtm->rtm_type = RTN_UNREACHABLE;
2084 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2085 rtm->rtm_type = RTN_LOCAL;
2087 rtm->rtm_type = RTN_UNICAST;
2089 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2090 rtm->rtm_protocol = rt->rt6i_protocol;
2091 if (rt->rt6i_flags&RTF_DYNAMIC)
2092 rtm->rtm_protocol = RTPROT_REDIRECT;
2093 else if (rt->rt6i_flags & RTF_ADDRCONF)
2094 rtm->rtm_protocol = RTPROT_KERNEL;
2095 else if (rt->rt6i_flags&RTF_DEFAULT)
2096 rtm->rtm_protocol = RTPROT_RA;
2098 if (rt->rt6i_flags&RTF_CACHE)
2099 rtm->rtm_flags |= RTM_F_CLONED;
2102 NLA_PUT(skb, RTA_DST, 16, dst);
2103 rtm->rtm_dst_len = 128;
2104 } else if (rtm->rtm_dst_len)
2105 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2106 #ifdef CONFIG_IPV6_SUBTREES
2108 NLA_PUT(skb, RTA_SRC, 16, src);
2109 rtm->rtm_src_len = 128;
2110 } else if (rtm->rtm_src_len)
2111 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2114 NLA_PUT_U32(skb, RTA_IIF, iif);
2116 struct in6_addr saddr_buf;
2117 if (ipv6_get_saddr(&rt->u.dst, dst, &saddr_buf) == 0)
2118 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2121 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2122 goto nla_put_failure;
2124 if (rt->u.dst.neighbour)
2125 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2128 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2130 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2132 expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2133 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2134 expires, rt->u.dst.error) < 0)
2135 goto nla_put_failure;
2137 return nlmsg_end(skb, nlh);
2140 nlmsg_cancel(skb, nlh);
2144 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2146 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2149 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2150 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2151 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2155 return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2156 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2157 prefix, NLM_F_MULTI);
2160 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2162 struct net *net = in_skb->sk->sk_net;
2163 struct nlattr *tb[RTA_MAX+1];
2164 struct rt6_info *rt;
2165 struct sk_buff *skb;
2170 if (net != &init_net)
2173 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2178 memset(&fl, 0, sizeof(fl));
2181 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2184 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2188 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2191 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2195 iif = nla_get_u32(tb[RTA_IIF]);
2198 fl.oif = nla_get_u32(tb[RTA_OIF]);
2201 struct net_device *dev;
2202 dev = __dev_get_by_index(&init_net, iif);
2209 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2215 /* Reserve room for dummy headers, this skb can pass
2216 through good chunk of routing engine.
2218 skb_reset_mac_header(skb);
2219 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2221 rt = (struct rt6_info*) ip6_route_output(NULL, &fl);
2222 skb->dst = &rt->u.dst;
2224 err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2225 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2226 nlh->nlmsg_seq, 0, 0);
2232 err = rtnl_unicast(skb, &init_net, NETLINK_CB(in_skb).pid);
2237 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2239 struct sk_buff *skb;
2244 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2246 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2250 err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2251 event, info->pid, seq, 0, 0);
2253 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2254 WARN_ON(err == -EMSGSIZE);
2258 err = rtnl_notify(skb, &init_net, info->pid,
2259 RTNLGRP_IPV6_ROUTE, info->nlh, gfp_any());
2262 rtnl_set_sk_err(&init_net, RTNLGRP_IPV6_ROUTE, err);
2269 #ifdef CONFIG_PROC_FS
2271 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2282 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2284 struct seq_file *m = p_arg;
2286 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2289 #ifdef CONFIG_IPV6_SUBTREES
2290 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2293 seq_puts(m, "00000000000000000000000000000000 00 ");
2296 if (rt->rt6i_nexthop) {
2297 seq_printf(m, NIP6_SEQFMT,
2298 NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2300 seq_puts(m, "00000000000000000000000000000000");
2302 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2303 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2304 rt->u.dst.__use, rt->rt6i_flags,
2305 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2309 static int ipv6_route_show(struct seq_file *m, void *v)
2311 fib6_clean_all(rt6_info_route, 0, m);
2315 static int ipv6_route_open(struct inode *inode, struct file *file)
2317 return single_open(file, ipv6_route_show, NULL);
2320 static const struct file_operations ipv6_route_proc_fops = {
2321 .owner = THIS_MODULE,
2322 .open = ipv6_route_open,
2324 .llseek = seq_lseek,
2325 .release = single_release,
2328 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2330 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2331 rt6_stats.fib_nodes, rt6_stats.fib_route_nodes,
2332 rt6_stats.fib_rt_alloc, rt6_stats.fib_rt_entries,
2333 rt6_stats.fib_rt_cache,
2334 atomic_read(&ip6_dst_ops.entries),
2335 rt6_stats.fib_discarded_routes);
2340 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2342 return single_open(file, rt6_stats_seq_show, NULL);
2345 static const struct file_operations rt6_stats_seq_fops = {
2346 .owner = THIS_MODULE,
2347 .open = rt6_stats_seq_open,
2349 .llseek = seq_lseek,
2350 .release = single_release,
2353 static int ipv6_route_proc_init(struct net *net)
2356 if (!proc_net_fops_create(net, "ipv6_route",
2357 0, &ipv6_route_proc_fops))
2360 if (!proc_net_fops_create(net, "rt6_stats",
2361 S_IRUGO, &rt6_stats_seq_fops))
2362 goto out_ipv6_route;
2368 proc_net_remove(net, "ipv6_route");
2372 static void ipv6_route_proc_fini(struct net *net)
2374 proc_net_remove(net, "ipv6_route");
2375 proc_net_remove(net, "rt6_stats");
2378 static inline int ipv6_route_proc_init(struct net *net)
2382 static inline void ipv6_route_proc_fini(struct net *net)
2386 #endif /* CONFIG_PROC_FS */
2388 #ifdef CONFIG_SYSCTL
2391 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2392 void __user *buffer, size_t *lenp, loff_t *ppos)
2394 int delay = init_net.ipv6.sysctl.flush_delay;
2396 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2397 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay);
2403 ctl_table ipv6_route_table_template[] = {
2405 .procname = "flush",
2406 .data = &init_net.ipv6.sysctl.flush_delay,
2407 .maxlen = sizeof(int),
2409 .proc_handler = &ipv6_sysctl_rtcache_flush
2412 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2413 .procname = "gc_thresh",
2414 .data = &ip6_dst_ops.gc_thresh,
2415 .maxlen = sizeof(int),
2417 .proc_handler = &proc_dointvec,
2420 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2421 .procname = "max_size",
2422 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2423 .maxlen = sizeof(int),
2425 .proc_handler = &proc_dointvec,
2428 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2429 .procname = "gc_min_interval",
2430 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2431 .maxlen = sizeof(int),
2433 .proc_handler = &proc_dointvec_jiffies,
2434 .strategy = &sysctl_jiffies,
2437 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2438 .procname = "gc_timeout",
2439 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2440 .maxlen = sizeof(int),
2442 .proc_handler = &proc_dointvec_jiffies,
2443 .strategy = &sysctl_jiffies,
2446 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2447 .procname = "gc_interval",
2448 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2449 .maxlen = sizeof(int),
2451 .proc_handler = &proc_dointvec_jiffies,
2452 .strategy = &sysctl_jiffies,
2455 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2456 .procname = "gc_elasticity",
2457 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2458 .maxlen = sizeof(int),
2460 .proc_handler = &proc_dointvec_jiffies,
2461 .strategy = &sysctl_jiffies,
2464 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2465 .procname = "mtu_expires",
2466 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2467 .maxlen = sizeof(int),
2469 .proc_handler = &proc_dointvec_jiffies,
2470 .strategy = &sysctl_jiffies,
2473 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2474 .procname = "min_adv_mss",
2475 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2476 .maxlen = sizeof(int),
2478 .proc_handler = &proc_dointvec_jiffies,
2479 .strategy = &sysctl_jiffies,
2482 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2483 .procname = "gc_min_interval_ms",
2484 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2485 .maxlen = sizeof(int),
2487 .proc_handler = &proc_dointvec_ms_jiffies,
2488 .strategy = &sysctl_ms_jiffies,
2493 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2495 struct ctl_table *table;
2497 table = kmemdup(ipv6_route_table_template,
2498 sizeof(ipv6_route_table_template),
2504 int __init ip6_route_init(void)
2508 ip6_dst_ops.kmem_cachep =
2509 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2510 SLAB_HWCACHE_ALIGN, NULL);
2511 if (!ip6_dst_ops.kmem_cachep)
2514 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops.kmem_cachep;
2518 goto out_kmem_cache;
2520 ret = ipv6_route_proc_init(&init_net);
2528 ret = fib6_rules_init();
2533 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2534 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2535 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2536 goto fib6_rules_init;
2543 fib6_rules_cleanup();
2547 ipv6_route_proc_fini(&init_net);
2552 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);
2556 void ip6_route_cleanup(void)
2558 fib6_rules_cleanup();
2559 ipv6_route_proc_fini(&init_net);
2563 kmem_cache_destroy(ip6_dst_ops.kmem_cachep);