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 <linux/nsproxy.h>
44 #include <net/net_namespace.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
58 #include <asm/uaccess.h>
61 #include <linux/sysctl.h>
64 /* Set to 3 to get tracing. */
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
72 #define RT6_TRACE(x...) do { ; } while (0)
75 #define CLONE_OFFLINK_ROUTE 0
77 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
78 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
79 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
80 static void ip6_dst_destroy(struct dst_entry *);
81 static void ip6_dst_ifdown(struct dst_entry *,
82 struct net_device *dev, int how);
83 static int ip6_dst_gc(struct dst_ops *ops);
85 static int ip6_pkt_discard(struct sk_buff *skb);
86 static int ip6_pkt_discard_out(struct sk_buff *skb);
87 static void ip6_link_failure(struct sk_buff *skb);
88 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
90 #ifdef CONFIG_IPV6_ROUTE_INFO
91 static struct rt6_info *rt6_add_route_info(struct net *net,
92 struct in6_addr *prefix, int prefixlen,
93 struct in6_addr *gwaddr, int ifindex,
95 static struct rt6_info *rt6_get_route_info(struct net *net,
96 struct in6_addr *prefix, int prefixlen,
97 struct in6_addr *gwaddr, int ifindex);
100 static struct dst_ops ip6_dst_ops_template = {
102 .protocol = __constant_htons(ETH_P_IPV6),
105 .check = ip6_dst_check,
106 .destroy = ip6_dst_destroy,
107 .ifdown = ip6_dst_ifdown,
108 .negative_advice = ip6_negative_advice,
109 .link_failure = ip6_link_failure,
110 .update_pmtu = ip6_rt_update_pmtu,
111 .local_out = ip6_local_out,
112 .entry_size = sizeof(struct rt6_info),
113 .entries = ATOMIC_INIT(0),
116 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
120 static struct dst_ops ip6_dst_blackhole_ops = {
122 .protocol = __constant_htons(ETH_P_IPV6),
123 .destroy = ip6_dst_destroy,
124 .check = ip6_dst_check,
125 .update_pmtu = ip6_rt_blackhole_update_pmtu,
126 .entry_size = sizeof(struct rt6_info),
127 .entries = ATOMIC_INIT(0),
130 static struct rt6_info ip6_null_entry_template = {
133 .__refcnt = ATOMIC_INIT(1),
136 .error = -ENETUNREACH,
137 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
138 .input = ip6_pkt_discard,
139 .output = ip6_pkt_discard_out,
142 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
143 .rt6i_metric = ~(u32) 0,
144 .rt6i_ref = ATOMIC_INIT(1),
147 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
149 static int ip6_pkt_prohibit(struct sk_buff *skb);
150 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
152 struct rt6_info ip6_prohibit_entry_template = {
155 .__refcnt = ATOMIC_INIT(1),
159 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
160 .input = ip6_pkt_prohibit,
161 .output = ip6_pkt_prohibit_out,
164 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
165 .rt6i_metric = ~(u32) 0,
166 .rt6i_ref = ATOMIC_INIT(1),
169 static struct rt6_info ip6_blk_hole_entry_template = {
172 .__refcnt = ATOMIC_INIT(1),
176 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
177 .input = dst_discard,
178 .output = dst_discard,
181 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
182 .rt6i_metric = ~(u32) 0,
183 .rt6i_ref = ATOMIC_INIT(1),
188 /* allocate dst with ip6_dst_ops */
189 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
191 return (struct rt6_info *)dst_alloc(ops);
194 static void ip6_dst_destroy(struct dst_entry *dst)
196 struct rt6_info *rt = (struct rt6_info *)dst;
197 struct inet6_dev *idev = rt->rt6i_idev;
200 rt->rt6i_idev = NULL;
205 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
208 struct rt6_info *rt = (struct rt6_info *)dst;
209 struct inet6_dev *idev = rt->rt6i_idev;
210 struct net_device *loopback_dev =
211 dev->nd_net->loopback_dev;
213 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
214 struct inet6_dev *loopback_idev =
215 in6_dev_get(loopback_dev);
216 if (loopback_idev != NULL) {
217 rt->rt6i_idev = loopback_idev;
223 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
225 return (rt->rt6i_flags & RTF_EXPIRES &&
226 time_after(jiffies, rt->rt6i_expires));
229 static inline int rt6_need_strict(struct in6_addr *daddr)
231 return (ipv6_addr_type(daddr) &
232 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
236 * Route lookup. Any table->tb6_lock is implied.
239 static inline struct rt6_info *rt6_device_match(struct net *net,
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 net->ipv6.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;
412 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
413 __func__, fn->leaf, oif);
417 fn->rr_ptr = rt0 = fn->leaf;
419 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
422 (strict & RT6_LOOKUP_F_REACHABLE)) {
423 struct rt6_info *next = rt0->u.dst.rt6_next;
425 /* no entries matched; do round-robin */
426 if (!next || next->rt6i_metric != rt0->rt6i_metric)
433 RT6_TRACE("%s() => %p\n",
436 net = rt0->rt6i_dev->nd_net;
437 return (match ? match : net->ipv6.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 net *net = dev->nd_net;
445 struct route_info *rinfo = (struct route_info *) opt;
446 struct in6_addr prefix_buf, *prefix;
451 if (len < sizeof(struct route_info)) {
455 /* Sanity check for prefix_len and length */
456 if (rinfo->length > 3) {
458 } else if (rinfo->prefix_len > 128) {
460 } else if (rinfo->prefix_len > 64) {
461 if (rinfo->length < 2) {
464 } else if (rinfo->prefix_len > 0) {
465 if (rinfo->length < 1) {
470 pref = rinfo->route_pref;
471 if (pref == ICMPV6_ROUTER_PREF_INVALID)
472 pref = ICMPV6_ROUTER_PREF_MEDIUM;
474 lifetime = ntohl(rinfo->lifetime);
475 if (lifetime == 0xffffffff) {
477 } else if (lifetime > 0x7fffffff/HZ) {
478 /* Avoid arithmetic overflow */
479 lifetime = 0x7fffffff/HZ - 1;
482 if (rinfo->length == 3)
483 prefix = (struct in6_addr *)rinfo->prefix;
485 /* this function is safe */
486 ipv6_addr_prefix(&prefix_buf,
487 (struct in6_addr *)rinfo->prefix,
489 prefix = &prefix_buf;
492 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
495 if (rt && !lifetime) {
501 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
504 rt->rt6i_flags = RTF_ROUTEINFO |
505 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
508 if (lifetime == 0xffffffff) {
509 rt->rt6i_flags &= ~RTF_EXPIRES;
511 rt->rt6i_expires = jiffies + HZ * lifetime;
512 rt->rt6i_flags |= RTF_EXPIRES;
514 dst_release(&rt->u.dst);
520 #define BACKTRACK(__net, saddr) \
522 if (rt == __net->ipv6.ip6_null_entry) { \
523 struct fib6_node *pn; \
525 if (fn->fn_flags & RTN_TL_ROOT) \
528 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
529 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
532 if (fn->fn_flags & RTN_RTINFO) \
538 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
539 struct fib6_table *table,
540 struct flowi *fl, int flags)
542 struct fib6_node *fn;
545 read_lock_bh(&table->tb6_lock);
546 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
549 rt = rt6_device_match(net, rt, fl->oif, flags);
550 BACKTRACK(net, &fl->fl6_src);
552 dst_use(&rt->u.dst, jiffies);
553 read_unlock_bh(&table->tb6_lock);
558 struct rt6_info *rt6_lookup(struct net *net, struct in6_addr *daddr,
559 struct in6_addr *saddr, int oif, int strict)
569 struct dst_entry *dst;
570 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
573 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
574 flags |= RT6_LOOKUP_F_HAS_SADDR;
577 dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
579 return (struct rt6_info *) dst;
586 EXPORT_SYMBOL(rt6_lookup);
588 /* ip6_ins_rt is called with FREE table->tb6_lock.
589 It takes new route entry, the addition fails by any reason the
590 route is freed. In any case, if caller does not hold it, it may
594 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
597 struct fib6_table *table;
599 table = rt->rt6i_table;
600 write_lock_bh(&table->tb6_lock);
601 err = fib6_add(&table->tb6_root, rt, info);
602 write_unlock_bh(&table->tb6_lock);
607 int ip6_ins_rt(struct rt6_info *rt)
609 struct nl_info info = {
610 .nl_net = rt->rt6i_dev->nd_net,
612 return __ip6_ins_rt(rt, &info);
615 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
616 struct in6_addr *saddr)
624 rt = ip6_rt_copy(ort);
627 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
628 if (rt->rt6i_dst.plen != 128 &&
629 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
630 rt->rt6i_flags |= RTF_ANYCAST;
631 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
634 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
635 rt->rt6i_dst.plen = 128;
636 rt->rt6i_flags |= RTF_CACHE;
637 rt->u.dst.flags |= DST_HOST;
639 #ifdef CONFIG_IPV6_SUBTREES
640 if (rt->rt6i_src.plen && saddr) {
641 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
642 rt->rt6i_src.plen = 128;
646 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
653 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
655 struct rt6_info *rt = ip6_rt_copy(ort);
657 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
658 rt->rt6i_dst.plen = 128;
659 rt->rt6i_flags |= RTF_CACHE;
660 rt->u.dst.flags |= DST_HOST;
661 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
666 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
667 struct flowi *fl, int flags)
669 struct fib6_node *fn;
670 struct rt6_info *rt, *nrt;
674 int reachable = ipv6_devconf.forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
676 strict |= flags & RT6_LOOKUP_F_IFACE;
679 read_lock_bh(&table->tb6_lock);
682 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
685 rt = rt6_select(fn, oif, strict | reachable);
687 BACKTRACK(net, &fl->fl6_src);
688 if (rt == net->ipv6.ip6_null_entry ||
689 rt->rt6i_flags & RTF_CACHE)
692 dst_hold(&rt->u.dst);
693 read_unlock_bh(&table->tb6_lock);
695 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
696 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
698 #if CLONE_OFFLINK_ROUTE
699 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
705 dst_release(&rt->u.dst);
706 rt = nrt ? : net->ipv6.ip6_null_entry;
708 dst_hold(&rt->u.dst);
710 err = ip6_ins_rt(nrt);
719 * Race condition! In the gap, when table->tb6_lock was
720 * released someone could insert this route. Relookup.
722 dst_release(&rt->u.dst);
730 dst_hold(&rt->u.dst);
731 read_unlock_bh(&table->tb6_lock);
733 rt->u.dst.lastuse = jiffies;
739 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
740 struct flowi *fl, int flags)
742 return ip6_pol_route(net, table, fl->iif, fl, flags);
745 void ip6_route_input(struct sk_buff *skb)
747 struct ipv6hdr *iph = ipv6_hdr(skb);
748 struct net *net = skb->dev->nd_net;
749 int flags = RT6_LOOKUP_F_HAS_SADDR;
751 .iif = skb->dev->ifindex,
756 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
760 .proto = iph->nexthdr,
763 if (rt6_need_strict(&iph->daddr))
764 flags |= RT6_LOOKUP_F_IFACE;
766 skb->dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input);
769 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
770 struct flowi *fl, int flags)
772 return ip6_pol_route(net, table, fl->oif, fl, flags);
775 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
780 if (rt6_need_strict(&fl->fl6_dst))
781 flags |= RT6_LOOKUP_F_IFACE;
783 if (!ipv6_addr_any(&fl->fl6_src))
784 flags |= RT6_LOOKUP_F_HAS_SADDR;
786 unsigned int prefs = inet6_sk(sk)->srcprefs;
787 if (prefs & IPV6_PREFER_SRC_TMP)
788 flags |= RT6_LOOKUP_F_SRCPREF_TMP;
789 if (prefs & IPV6_PREFER_SRC_PUBLIC)
790 flags |= RT6_LOOKUP_F_SRCPREF_PUBLIC;
791 if (prefs & IPV6_PREFER_SRC_COA)
792 flags |= RT6_LOOKUP_F_SRCPREF_COA;
795 return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
798 EXPORT_SYMBOL(ip6_route_output);
800 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
802 struct rt6_info *ort = (struct rt6_info *) *dstp;
803 struct rt6_info *rt = (struct rt6_info *)
804 dst_alloc(&ip6_dst_blackhole_ops);
805 struct dst_entry *new = NULL;
810 atomic_set(&new->__refcnt, 1);
812 new->input = dst_discard;
813 new->output = dst_discard;
815 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
816 new->dev = ort->u.dst.dev;
819 rt->rt6i_idev = ort->rt6i_idev;
821 in6_dev_hold(rt->rt6i_idev);
822 rt->rt6i_expires = 0;
824 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
825 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
828 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
829 #ifdef CONFIG_IPV6_SUBTREES
830 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
838 return (new ? 0 : -ENOMEM);
840 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
843 * Destination cache support functions
846 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
850 rt = (struct rt6_info *) dst;
852 if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
858 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
860 struct rt6_info *rt = (struct rt6_info *) dst;
863 if (rt->rt6i_flags & RTF_CACHE)
871 static void ip6_link_failure(struct sk_buff *skb)
875 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
877 rt = (struct rt6_info *) skb->dst;
879 if (rt->rt6i_flags&RTF_CACHE) {
880 dst_set_expires(&rt->u.dst, 0);
881 rt->rt6i_flags |= RTF_EXPIRES;
882 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
883 rt->rt6i_node->fn_sernum = -1;
887 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
889 struct rt6_info *rt6 = (struct rt6_info*)dst;
891 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
892 rt6->rt6i_flags |= RTF_MODIFIED;
893 if (mtu < IPV6_MIN_MTU) {
895 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
897 dst->metrics[RTAX_MTU-1] = mtu;
898 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
902 static int ipv6_get_mtu(struct net_device *dev);
904 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
906 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
908 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
909 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
912 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
913 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
914 * IPV6_MAXPLEN is also valid and means: "any MSS,
915 * rely only on pmtu discovery"
917 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
922 static struct dst_entry *icmp6_dst_gc_list;
923 static DEFINE_SPINLOCK(icmp6_dst_lock);
925 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
926 struct neighbour *neigh,
927 struct in6_addr *addr)
930 struct inet6_dev *idev = in6_dev_get(dev);
931 struct net *net = dev->nd_net;
933 if (unlikely(idev == NULL))
936 rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
937 if (unlikely(rt == NULL)) {
946 neigh = ndisc_get_neigh(dev, addr);
949 rt->rt6i_idev = idev;
950 rt->rt6i_nexthop = neigh;
951 atomic_set(&rt->u.dst.__refcnt, 1);
952 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
953 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
954 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
955 rt->u.dst.output = ip6_output;
957 #if 0 /* there's no chance to use these for ndisc */
958 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
961 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
962 rt->rt6i_dst.plen = 128;
965 spin_lock_bh(&icmp6_dst_lock);
966 rt->u.dst.next = icmp6_dst_gc_list;
967 icmp6_dst_gc_list = &rt->u.dst;
968 spin_unlock_bh(&icmp6_dst_lock);
970 fib6_force_start_gc(net);
976 int icmp6_dst_gc(int *more)
978 struct dst_entry *dst, *next, **pprev;
984 spin_lock_bh(&icmp6_dst_lock);
985 pprev = &icmp6_dst_gc_list;
987 while ((dst = *pprev) != NULL) {
988 if (!atomic_read(&dst->__refcnt)) {
998 spin_unlock_bh(&icmp6_dst_lock);
1003 static int ip6_dst_gc(struct dst_ops *ops)
1005 unsigned long now = jiffies;
1006 struct net *net = ops->dst_net;
1007 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1008 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1009 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1010 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1011 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1013 if (time_after(rt_last_gc + rt_min_interval, now) &&
1014 atomic_read(&ops->entries) <= rt_max_size)
1017 net->ipv6.ip6_rt_gc_expire++;
1018 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1019 net->ipv6.ip6_rt_last_gc = now;
1020 if (atomic_read(&ops->entries) < ops->gc_thresh)
1021 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1023 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1024 return (atomic_read(&ops->entries) > rt_max_size);
1027 /* Clean host part of a prefix. Not necessary in radix tree,
1028 but results in cleaner routing tables.
1030 Remove it only when all the things will work!
1033 static int ipv6_get_mtu(struct net_device *dev)
1035 int mtu = IPV6_MIN_MTU;
1036 struct inet6_dev *idev;
1038 idev = in6_dev_get(dev);
1040 mtu = idev->cnf.mtu6;
1046 int ip6_dst_hoplimit(struct dst_entry *dst)
1048 int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1050 struct net_device *dev = dst->dev;
1051 struct inet6_dev *idev = in6_dev_get(dev);
1053 hoplimit = idev->cnf.hop_limit;
1056 hoplimit = ipv6_devconf.hop_limit;
1065 int ip6_route_add(struct fib6_config *cfg)
1068 struct net *net = cfg->fc_nlinfo.nl_net;
1069 struct rt6_info *rt = NULL;
1070 struct net_device *dev = NULL;
1071 struct inet6_dev *idev = NULL;
1072 struct fib6_table *table;
1075 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1077 #ifndef CONFIG_IPV6_SUBTREES
1078 if (cfg->fc_src_len)
1081 if (cfg->fc_ifindex) {
1083 dev = dev_get_by_index(net, cfg->fc_ifindex);
1086 idev = in6_dev_get(dev);
1091 if (cfg->fc_metric == 0)
1092 cfg->fc_metric = IP6_RT_PRIO_USER;
1094 table = fib6_new_table(net, cfg->fc_table);
1095 if (table == NULL) {
1100 rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1107 rt->u.dst.obsolete = -1;
1108 rt->rt6i_expires = jiffies + clock_t_to_jiffies(cfg->fc_expires);
1110 if (cfg->fc_protocol == RTPROT_UNSPEC)
1111 cfg->fc_protocol = RTPROT_BOOT;
1112 rt->rt6i_protocol = cfg->fc_protocol;
1114 addr_type = ipv6_addr_type(&cfg->fc_dst);
1116 if (addr_type & IPV6_ADDR_MULTICAST)
1117 rt->u.dst.input = ip6_mc_input;
1119 rt->u.dst.input = ip6_forward;
1121 rt->u.dst.output = ip6_output;
1123 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1124 rt->rt6i_dst.plen = cfg->fc_dst_len;
1125 if (rt->rt6i_dst.plen == 128)
1126 rt->u.dst.flags = DST_HOST;
1128 #ifdef CONFIG_IPV6_SUBTREES
1129 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1130 rt->rt6i_src.plen = cfg->fc_src_len;
1133 rt->rt6i_metric = cfg->fc_metric;
1135 /* We cannot add true routes via loopback here,
1136 they would result in kernel looping; promote them to reject routes
1138 if ((cfg->fc_flags & RTF_REJECT) ||
1139 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1140 /* hold loopback dev/idev if we haven't done so. */
1141 if (dev != net->loopback_dev) {
1146 dev = net->loopback_dev;
1148 idev = in6_dev_get(dev);
1154 rt->u.dst.output = ip6_pkt_discard_out;
1155 rt->u.dst.input = ip6_pkt_discard;
1156 rt->u.dst.error = -ENETUNREACH;
1157 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1161 if (cfg->fc_flags & RTF_GATEWAY) {
1162 struct in6_addr *gw_addr;
1165 gw_addr = &cfg->fc_gateway;
1166 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1167 gwa_type = ipv6_addr_type(gw_addr);
1169 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1170 struct rt6_info *grt;
1172 /* IPv6 strictly inhibits using not link-local
1173 addresses as nexthop address.
1174 Otherwise, router will not able to send redirects.
1175 It is very good, but in some (rare!) circumstances
1176 (SIT, PtP, NBMA NOARP links) it is handy to allow
1177 some exceptions. --ANK
1180 if (!(gwa_type&IPV6_ADDR_UNICAST))
1183 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1185 err = -EHOSTUNREACH;
1189 if (dev != grt->rt6i_dev) {
1190 dst_release(&grt->u.dst);
1194 dev = grt->rt6i_dev;
1195 idev = grt->rt6i_idev;
1197 in6_dev_hold(grt->rt6i_idev);
1199 if (!(grt->rt6i_flags&RTF_GATEWAY))
1201 dst_release(&grt->u.dst);
1207 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1215 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1216 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1217 if (IS_ERR(rt->rt6i_nexthop)) {
1218 err = PTR_ERR(rt->rt6i_nexthop);
1219 rt->rt6i_nexthop = NULL;
1224 rt->rt6i_flags = cfg->fc_flags;
1231 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1232 int type = nla_type(nla);
1235 if (type > RTAX_MAX) {
1240 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1245 if (rt->u.dst.metrics[RTAX_HOPLIMIT-1] == 0)
1246 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1247 if (!rt->u.dst.metrics[RTAX_MTU-1])
1248 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1249 if (!rt->u.dst.metrics[RTAX_ADVMSS-1])
1250 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1251 rt->u.dst.dev = dev;
1252 rt->rt6i_idev = idev;
1253 rt->rt6i_table = table;
1255 cfg->fc_nlinfo.nl_net = dev->nd_net;
1257 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1265 dst_free(&rt->u.dst);
1269 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1272 struct fib6_table *table;
1273 struct net *net = rt->rt6i_dev->nd_net;
1275 if (rt == net->ipv6.ip6_null_entry)
1278 table = rt->rt6i_table;
1279 write_lock_bh(&table->tb6_lock);
1281 err = fib6_del(rt, info);
1282 dst_release(&rt->u.dst);
1284 write_unlock_bh(&table->tb6_lock);
1289 int ip6_del_rt(struct rt6_info *rt)
1291 struct nl_info info = {
1292 .nl_net = rt->rt6i_dev->nd_net,
1294 return __ip6_del_rt(rt, &info);
1297 static int ip6_route_del(struct fib6_config *cfg)
1299 struct fib6_table *table;
1300 struct fib6_node *fn;
1301 struct rt6_info *rt;
1304 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1308 read_lock_bh(&table->tb6_lock);
1310 fn = fib6_locate(&table->tb6_root,
1311 &cfg->fc_dst, cfg->fc_dst_len,
1312 &cfg->fc_src, cfg->fc_src_len);
1315 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1316 if (cfg->fc_ifindex &&
1317 (rt->rt6i_dev == NULL ||
1318 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1320 if (cfg->fc_flags & RTF_GATEWAY &&
1321 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1323 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1325 dst_hold(&rt->u.dst);
1326 read_unlock_bh(&table->tb6_lock);
1328 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1331 read_unlock_bh(&table->tb6_lock);
1339 struct ip6rd_flowi {
1341 struct in6_addr gateway;
1344 static struct rt6_info *__ip6_route_redirect(struct net *net,
1345 struct fib6_table *table,
1349 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1350 struct rt6_info *rt;
1351 struct fib6_node *fn;
1354 * Get the "current" route for this destination and
1355 * check if the redirect has come from approriate router.
1357 * RFC 2461 specifies that redirects should only be
1358 * accepted if they come from the nexthop to the target.
1359 * Due to the way the routes are chosen, this notion
1360 * is a bit fuzzy and one might need to check all possible
1364 read_lock_bh(&table->tb6_lock);
1365 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1367 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1369 * Current route is on-link; redirect is always invalid.
1371 * Seems, previous statement is not true. It could
1372 * be node, which looks for us as on-link (f.e. proxy ndisc)
1373 * But then router serving it might decide, that we should
1374 * know truth 8)8) --ANK (980726).
1376 if (rt6_check_expired(rt))
1378 if (!(rt->rt6i_flags & RTF_GATEWAY))
1380 if (fl->oif != rt->rt6i_dev->ifindex)
1382 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1388 rt = net->ipv6.ip6_null_entry;
1389 BACKTRACK(net, &fl->fl6_src);
1391 dst_hold(&rt->u.dst);
1393 read_unlock_bh(&table->tb6_lock);
1398 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1399 struct in6_addr *src,
1400 struct in6_addr *gateway,
1401 struct net_device *dev)
1403 int flags = RT6_LOOKUP_F_HAS_SADDR;
1404 struct net *net = dev->nd_net;
1405 struct ip6rd_flowi rdfl = {
1407 .oif = dev->ifindex,
1415 .gateway = *gateway,
1418 if (rt6_need_strict(dest))
1419 flags |= RT6_LOOKUP_F_IFACE;
1421 return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1422 flags, __ip6_route_redirect);
1425 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1426 struct in6_addr *saddr,
1427 struct neighbour *neigh, u8 *lladdr, int on_link)
1429 struct rt6_info *rt, *nrt = NULL;
1430 struct netevent_redirect netevent;
1431 struct net *net = neigh->dev->nd_net;
1433 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1435 if (rt == net->ipv6.ip6_null_entry) {
1436 if (net_ratelimit())
1437 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1438 "for redirect target\n");
1443 * We have finally decided to accept it.
1446 neigh_update(neigh, lladdr, NUD_STALE,
1447 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1448 NEIGH_UPDATE_F_OVERRIDE|
1449 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1450 NEIGH_UPDATE_F_ISROUTER))
1454 * Redirect received -> path was valid.
1455 * Look, redirects are sent only in response to data packets,
1456 * so that this nexthop apparently is reachable. --ANK
1458 dst_confirm(&rt->u.dst);
1460 /* Duplicate redirect: silently ignore. */
1461 if (neigh == rt->u.dst.neighbour)
1464 nrt = ip6_rt_copy(rt);
1468 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1470 nrt->rt6i_flags &= ~RTF_GATEWAY;
1472 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1473 nrt->rt6i_dst.plen = 128;
1474 nrt->u.dst.flags |= DST_HOST;
1476 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1477 nrt->rt6i_nexthop = neigh_clone(neigh);
1478 /* Reset pmtu, it may be better */
1479 nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1480 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(neigh->dev->nd_net,
1481 dst_mtu(&nrt->u.dst));
1483 if (ip6_ins_rt(nrt))
1486 netevent.old = &rt->u.dst;
1487 netevent.new = &nrt->u.dst;
1488 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1490 if (rt->rt6i_flags&RTF_CACHE) {
1496 dst_release(&rt->u.dst);
1501 * Handle ICMP "packet too big" messages
1502 * i.e. Path MTU discovery
1505 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1506 struct net_device *dev, u32 pmtu)
1508 struct rt6_info *rt, *nrt;
1509 struct net *net = dev->nd_net;
1512 rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1516 if (pmtu >= dst_mtu(&rt->u.dst))
1519 if (pmtu < IPV6_MIN_MTU) {
1521 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1522 * MTU (1280) and a fragment header should always be included
1523 * after a node receiving Too Big message reporting PMTU is
1524 * less than the IPv6 Minimum Link MTU.
1526 pmtu = IPV6_MIN_MTU;
1530 /* New mtu received -> path was valid.
1531 They are sent only in response to data packets,
1532 so that this nexthop apparently is reachable. --ANK
1534 dst_confirm(&rt->u.dst);
1536 /* Host route. If it is static, it would be better
1537 not to override it, but add new one, so that
1538 when cache entry will expire old pmtu
1539 would return automatically.
1541 if (rt->rt6i_flags & RTF_CACHE) {
1542 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1544 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1545 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1546 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1551 Two cases are possible:
1552 1. It is connected route. Action: COW
1553 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1555 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1556 nrt = rt6_alloc_cow(rt, daddr, saddr);
1558 nrt = rt6_alloc_clone(rt, daddr);
1561 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1563 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1565 /* According to RFC 1981, detecting PMTU increase shouldn't be
1566 * happened within 5 mins, the recommended timer is 10 mins.
1567 * Here this route expiration time is set to ip6_rt_mtu_expires
1568 * which is 10 mins. After 10 mins the decreased pmtu is expired
1569 * and detecting PMTU increase will be automatically happened.
1571 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1572 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1577 dst_release(&rt->u.dst);
1581 * Misc support functions
1584 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1586 struct net *net = ort->rt6i_dev->nd_net;
1587 struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1590 rt->u.dst.input = ort->u.dst.input;
1591 rt->u.dst.output = ort->u.dst.output;
1593 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1594 rt->u.dst.error = ort->u.dst.error;
1595 rt->u.dst.dev = ort->u.dst.dev;
1597 dev_hold(rt->u.dst.dev);
1598 rt->rt6i_idev = ort->rt6i_idev;
1600 in6_dev_hold(rt->rt6i_idev);
1601 rt->u.dst.lastuse = jiffies;
1602 rt->rt6i_expires = 0;
1604 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1605 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1606 rt->rt6i_metric = 0;
1608 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1609 #ifdef CONFIG_IPV6_SUBTREES
1610 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1612 rt->rt6i_table = ort->rt6i_table;
1617 #ifdef CONFIG_IPV6_ROUTE_INFO
1618 static struct rt6_info *rt6_get_route_info(struct net *net,
1619 struct in6_addr *prefix, int prefixlen,
1620 struct in6_addr *gwaddr, int ifindex)
1622 struct fib6_node *fn;
1623 struct rt6_info *rt = NULL;
1624 struct fib6_table *table;
1626 table = fib6_get_table(net, RT6_TABLE_INFO);
1630 write_lock_bh(&table->tb6_lock);
1631 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1635 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1636 if (rt->rt6i_dev->ifindex != ifindex)
1638 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1640 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1642 dst_hold(&rt->u.dst);
1646 write_unlock_bh(&table->tb6_lock);
1650 static struct rt6_info *rt6_add_route_info(struct net *net,
1651 struct in6_addr *prefix, int prefixlen,
1652 struct in6_addr *gwaddr, int ifindex,
1655 struct fib6_config cfg = {
1656 .fc_table = RT6_TABLE_INFO,
1657 .fc_metric = IP6_RT_PRIO_USER,
1658 .fc_ifindex = ifindex,
1659 .fc_dst_len = prefixlen,
1660 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1661 RTF_UP | RTF_PREF(pref),
1663 .fc_nlinfo.nlh = NULL,
1664 .fc_nlinfo.nl_net = net,
1667 ipv6_addr_copy(&cfg.fc_dst, prefix);
1668 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1670 /* We should treat it as a default route if prefix length is 0. */
1672 cfg.fc_flags |= RTF_DEFAULT;
1674 ip6_route_add(&cfg);
1676 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1680 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1682 struct rt6_info *rt;
1683 struct fib6_table *table;
1685 table = fib6_get_table(dev->nd_net, RT6_TABLE_DFLT);
1689 write_lock_bh(&table->tb6_lock);
1690 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1691 if (dev == rt->rt6i_dev &&
1692 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1693 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1697 dst_hold(&rt->u.dst);
1698 write_unlock_bh(&table->tb6_lock);
1702 EXPORT_SYMBOL(rt6_get_dflt_router);
1704 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1705 struct net_device *dev,
1708 struct fib6_config cfg = {
1709 .fc_table = RT6_TABLE_DFLT,
1710 .fc_metric = IP6_RT_PRIO_USER,
1711 .fc_ifindex = dev->ifindex,
1712 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1713 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1715 .fc_nlinfo.nlh = NULL,
1716 .fc_nlinfo.nl_net = dev->nd_net,
1719 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1721 ip6_route_add(&cfg);
1723 return rt6_get_dflt_router(gwaddr, dev);
1726 void rt6_purge_dflt_routers(struct net *net)
1728 struct rt6_info *rt;
1729 struct fib6_table *table;
1731 /* NOTE: Keep consistent with rt6_get_dflt_router */
1732 table = fib6_get_table(net, RT6_TABLE_DFLT);
1737 read_lock_bh(&table->tb6_lock);
1738 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1739 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1740 dst_hold(&rt->u.dst);
1741 read_unlock_bh(&table->tb6_lock);
1746 read_unlock_bh(&table->tb6_lock);
1749 static void rtmsg_to_fib6_config(struct net *net,
1750 struct in6_rtmsg *rtmsg,
1751 struct fib6_config *cfg)
1753 memset(cfg, 0, sizeof(*cfg));
1755 cfg->fc_table = RT6_TABLE_MAIN;
1756 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1757 cfg->fc_metric = rtmsg->rtmsg_metric;
1758 cfg->fc_expires = rtmsg->rtmsg_info;
1759 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1760 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1761 cfg->fc_flags = rtmsg->rtmsg_flags;
1763 cfg->fc_nlinfo.nl_net = net;
1765 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1766 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1767 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1770 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1772 struct fib6_config cfg;
1773 struct in6_rtmsg rtmsg;
1777 case SIOCADDRT: /* Add a route */
1778 case SIOCDELRT: /* Delete a route */
1779 if (!capable(CAP_NET_ADMIN))
1781 err = copy_from_user(&rtmsg, arg,
1782 sizeof(struct in6_rtmsg));
1786 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1791 err = ip6_route_add(&cfg);
1794 err = ip6_route_del(&cfg);
1808 * Drop the packet on the floor
1811 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1814 switch (ipstats_mib_noroutes) {
1815 case IPSTATS_MIB_INNOROUTES:
1816 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1817 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1818 IP6_INC_STATS(ip6_dst_idev(skb->dst), IPSTATS_MIB_INADDRERRORS);
1822 case IPSTATS_MIB_OUTNOROUTES:
1823 IP6_INC_STATS(ip6_dst_idev(skb->dst), ipstats_mib_noroutes);
1826 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1831 static int ip6_pkt_discard(struct sk_buff *skb)
1833 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1836 static int ip6_pkt_discard_out(struct sk_buff *skb)
1838 skb->dev = skb->dst->dev;
1839 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1842 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1844 static int ip6_pkt_prohibit(struct sk_buff *skb)
1846 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1849 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1851 skb->dev = skb->dst->dev;
1852 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1858 * Allocate a dst for local (unicast / anycast) address.
1861 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1862 const struct in6_addr *addr,
1865 struct net *net = idev->dev->nd_net;
1866 struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1869 return ERR_PTR(-ENOMEM);
1871 dev_hold(net->loopback_dev);
1874 rt->u.dst.flags = DST_HOST;
1875 rt->u.dst.input = ip6_input;
1876 rt->u.dst.output = ip6_output;
1877 rt->rt6i_dev = net->loopback_dev;
1878 rt->rt6i_idev = idev;
1879 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1880 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1881 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1882 rt->u.dst.obsolete = -1;
1884 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1886 rt->rt6i_flags |= RTF_ANYCAST;
1888 rt->rt6i_flags |= RTF_LOCAL;
1889 rt->rt6i_nexthop = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1890 if (rt->rt6i_nexthop == NULL) {
1891 dst_free(&rt->u.dst);
1892 return ERR_PTR(-ENOMEM);
1895 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1896 rt->rt6i_dst.plen = 128;
1897 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1899 atomic_set(&rt->u.dst.__refcnt, 1);
1904 struct arg_dev_net {
1905 struct net_device *dev;
1909 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1911 struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1912 struct net *net = ((struct arg_dev_net *)arg)->net;
1914 if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1915 rt != net->ipv6.ip6_null_entry) {
1916 RT6_TRACE("deleted by ifdown %p\n", rt);
1922 void rt6_ifdown(struct net *net, struct net_device *dev)
1924 struct arg_dev_net adn = {
1929 fib6_clean_all(net, fib6_ifdown, 0, &adn);
1932 struct rt6_mtu_change_arg
1934 struct net_device *dev;
1938 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
1940 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
1941 struct inet6_dev *idev;
1942 struct net *net = arg->dev->nd_net;
1944 /* In IPv6 pmtu discovery is not optional,
1945 so that RTAX_MTU lock cannot disable it.
1946 We still use this lock to block changes
1947 caused by addrconf/ndisc.
1950 idev = __in6_dev_get(arg->dev);
1954 /* For administrative MTU increase, there is no way to discover
1955 IPv6 PMTU increase, so PMTU increase should be updated here.
1956 Since RFC 1981 doesn't include administrative MTU increase
1957 update PMTU increase is a MUST. (i.e. jumbo frame)
1960 If new MTU is less than route PMTU, this new MTU will be the
1961 lowest MTU in the path, update the route PMTU to reflect PMTU
1962 decreases; if new MTU is greater than route PMTU, and the
1963 old MTU is the lowest MTU in the path, update the route PMTU
1964 to reflect the increase. In this case if the other nodes' MTU
1965 also have the lowest MTU, TOO BIG MESSAGE will be lead to
1968 if (rt->rt6i_dev == arg->dev &&
1969 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
1970 (dst_mtu(&rt->u.dst) >= arg->mtu ||
1971 (dst_mtu(&rt->u.dst) < arg->mtu &&
1972 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
1973 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
1974 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
1979 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
1981 struct rt6_mtu_change_arg arg = {
1986 fib6_clean_all(dev->nd_net, rt6_mtu_change_route, 0, &arg);
1989 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
1990 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
1991 [RTA_OIF] = { .type = NLA_U32 },
1992 [RTA_IIF] = { .type = NLA_U32 },
1993 [RTA_PRIORITY] = { .type = NLA_U32 },
1994 [RTA_METRICS] = { .type = NLA_NESTED },
1997 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
1998 struct fib6_config *cfg)
2001 struct nlattr *tb[RTA_MAX+1];
2004 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2009 rtm = nlmsg_data(nlh);
2010 memset(cfg, 0, sizeof(*cfg));
2012 cfg->fc_table = rtm->rtm_table;
2013 cfg->fc_dst_len = rtm->rtm_dst_len;
2014 cfg->fc_src_len = rtm->rtm_src_len;
2015 cfg->fc_flags = RTF_UP;
2016 cfg->fc_protocol = rtm->rtm_protocol;
2018 if (rtm->rtm_type == RTN_UNREACHABLE)
2019 cfg->fc_flags |= RTF_REJECT;
2021 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2022 cfg->fc_nlinfo.nlh = nlh;
2023 cfg->fc_nlinfo.nl_net = skb->sk->sk_net;
2025 if (tb[RTA_GATEWAY]) {
2026 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2027 cfg->fc_flags |= RTF_GATEWAY;
2031 int plen = (rtm->rtm_dst_len + 7) >> 3;
2033 if (nla_len(tb[RTA_DST]) < plen)
2036 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2040 int plen = (rtm->rtm_src_len + 7) >> 3;
2042 if (nla_len(tb[RTA_SRC]) < plen)
2045 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2049 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2051 if (tb[RTA_PRIORITY])
2052 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2054 if (tb[RTA_METRICS]) {
2055 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2056 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2060 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2067 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2069 struct fib6_config cfg;
2072 err = rtm_to_fib6_config(skb, nlh, &cfg);
2076 return ip6_route_del(&cfg);
2079 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2081 struct fib6_config cfg;
2084 err = rtm_to_fib6_config(skb, nlh, &cfg);
2088 return ip6_route_add(&cfg);
2091 static inline size_t rt6_nlmsg_size(void)
2093 return NLMSG_ALIGN(sizeof(struct rtmsg))
2094 + nla_total_size(16) /* RTA_SRC */
2095 + nla_total_size(16) /* RTA_DST */
2096 + nla_total_size(16) /* RTA_GATEWAY */
2097 + nla_total_size(16) /* RTA_PREFSRC */
2098 + nla_total_size(4) /* RTA_TABLE */
2099 + nla_total_size(4) /* RTA_IIF */
2100 + nla_total_size(4) /* RTA_OIF */
2101 + nla_total_size(4) /* RTA_PRIORITY */
2102 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2103 + nla_total_size(sizeof(struct rta_cacheinfo));
2106 static int rt6_fill_node(struct sk_buff *skb, struct rt6_info *rt,
2107 struct in6_addr *dst, struct in6_addr *src,
2108 int iif, int type, u32 pid, u32 seq,
2109 int prefix, unsigned int flags)
2112 struct nlmsghdr *nlh;
2116 if (prefix) { /* user wants prefix routes only */
2117 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2118 /* success since this is not a prefix route */
2123 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2127 rtm = nlmsg_data(nlh);
2128 rtm->rtm_family = AF_INET6;
2129 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2130 rtm->rtm_src_len = rt->rt6i_src.plen;
2133 table = rt->rt6i_table->tb6_id;
2135 table = RT6_TABLE_UNSPEC;
2136 rtm->rtm_table = table;
2137 NLA_PUT_U32(skb, RTA_TABLE, table);
2138 if (rt->rt6i_flags&RTF_REJECT)
2139 rtm->rtm_type = RTN_UNREACHABLE;
2140 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2141 rtm->rtm_type = RTN_LOCAL;
2143 rtm->rtm_type = RTN_UNICAST;
2145 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2146 rtm->rtm_protocol = rt->rt6i_protocol;
2147 if (rt->rt6i_flags&RTF_DYNAMIC)
2148 rtm->rtm_protocol = RTPROT_REDIRECT;
2149 else if (rt->rt6i_flags & RTF_ADDRCONF)
2150 rtm->rtm_protocol = RTPROT_KERNEL;
2151 else if (rt->rt6i_flags&RTF_DEFAULT)
2152 rtm->rtm_protocol = RTPROT_RA;
2154 if (rt->rt6i_flags&RTF_CACHE)
2155 rtm->rtm_flags |= RTM_F_CLONED;
2158 NLA_PUT(skb, RTA_DST, 16, dst);
2159 rtm->rtm_dst_len = 128;
2160 } else if (rtm->rtm_dst_len)
2161 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2162 #ifdef CONFIG_IPV6_SUBTREES
2164 NLA_PUT(skb, RTA_SRC, 16, src);
2165 rtm->rtm_src_len = 128;
2166 } else if (rtm->rtm_src_len)
2167 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2170 NLA_PUT_U32(skb, RTA_IIF, iif);
2172 struct in6_addr saddr_buf;
2173 if (ipv6_dev_get_saddr(ip6_dst_idev(&rt->u.dst)->dev,
2174 dst, 0, &saddr_buf) == 0)
2175 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2178 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2179 goto nla_put_failure;
2181 if (rt->u.dst.neighbour)
2182 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2185 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2187 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2189 expires = rt->rt6i_expires ? rt->rt6i_expires - jiffies : 0;
2190 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2191 expires, rt->u.dst.error) < 0)
2192 goto nla_put_failure;
2194 return nlmsg_end(skb, nlh);
2197 nlmsg_cancel(skb, nlh);
2201 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2203 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2206 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2207 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2208 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2212 return rt6_fill_node(arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2213 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2214 prefix, NLM_F_MULTI);
2217 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2219 struct net *net = in_skb->sk->sk_net;
2220 struct nlattr *tb[RTA_MAX+1];
2221 struct rt6_info *rt;
2222 struct sk_buff *skb;
2227 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2232 memset(&fl, 0, sizeof(fl));
2235 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2238 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2242 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2245 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2249 iif = nla_get_u32(tb[RTA_IIF]);
2252 fl.oif = nla_get_u32(tb[RTA_OIF]);
2255 struct net_device *dev;
2256 dev = __dev_get_by_index(net, iif);
2263 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2269 /* Reserve room for dummy headers, this skb can pass
2270 through good chunk of routing engine.
2272 skb_reset_mac_header(skb);
2273 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2275 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2276 skb->dst = &rt->u.dst;
2278 err = rt6_fill_node(skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2279 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2280 nlh->nlmsg_seq, 0, 0);
2286 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2291 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2293 struct sk_buff *skb;
2294 struct net *net = info->nl_net;
2299 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2301 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2305 err = rt6_fill_node(skb, rt, NULL, NULL, 0,
2306 event, info->pid, seq, 0, 0);
2308 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2309 WARN_ON(err == -EMSGSIZE);
2313 err = rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2314 info->nlh, gfp_any());
2317 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2320 static int ip6_route_dev_notify(struct notifier_block *this,
2321 unsigned long event, void *data)
2323 struct net_device *dev = (struct net_device *)data;
2324 struct net *net = dev->nd_net;
2326 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2327 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2328 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2329 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2330 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2331 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2332 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2333 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2344 #ifdef CONFIG_PROC_FS
2346 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2357 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2359 struct seq_file *m = p_arg;
2361 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_dst.addr),
2364 #ifdef CONFIG_IPV6_SUBTREES
2365 seq_printf(m, NIP6_SEQFMT " %02x ", NIP6(rt->rt6i_src.addr),
2368 seq_puts(m, "00000000000000000000000000000000 00 ");
2371 if (rt->rt6i_nexthop) {
2372 seq_printf(m, NIP6_SEQFMT,
2373 NIP6(*((struct in6_addr *)rt->rt6i_nexthop->primary_key)));
2375 seq_puts(m, "00000000000000000000000000000000");
2377 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2378 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2379 rt->u.dst.__use, rt->rt6i_flags,
2380 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2384 static int ipv6_route_show(struct seq_file *m, void *v)
2386 struct net *net = (struct net *)m->private;
2387 fib6_clean_all(net, rt6_info_route, 0, m);
2391 static int ipv6_route_open(struct inode *inode, struct file *file)
2393 struct net *net = get_proc_net(inode);
2396 return single_open(file, ipv6_route_show, net);
2399 static int ipv6_route_release(struct inode *inode, struct file *file)
2401 struct seq_file *seq = file->private_data;
2402 struct net *net = seq->private;
2404 return single_release(inode, file);
2407 static const struct file_operations ipv6_route_proc_fops = {
2408 .owner = THIS_MODULE,
2409 .open = ipv6_route_open,
2411 .llseek = seq_lseek,
2412 .release = ipv6_route_release,
2415 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2417 struct net *net = (struct net *)seq->private;
2418 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2419 net->ipv6.rt6_stats->fib_nodes,
2420 net->ipv6.rt6_stats->fib_route_nodes,
2421 net->ipv6.rt6_stats->fib_rt_alloc,
2422 net->ipv6.rt6_stats->fib_rt_entries,
2423 net->ipv6.rt6_stats->fib_rt_cache,
2424 atomic_read(&net->ipv6.ip6_dst_ops->entries),
2425 net->ipv6.rt6_stats->fib_discarded_routes);
2430 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2432 struct net *net = get_proc_net(inode);
2433 return single_open(file, rt6_stats_seq_show, net);
2436 static int rt6_stats_seq_release(struct inode *inode, struct file *file)
2438 struct seq_file *seq = file->private_data;
2439 struct net *net = (struct net *)seq->private;
2441 return single_release(inode, file);
2444 static const struct file_operations rt6_stats_seq_fops = {
2445 .owner = THIS_MODULE,
2446 .open = rt6_stats_seq_open,
2448 .llseek = seq_lseek,
2449 .release = rt6_stats_seq_release,
2451 #endif /* CONFIG_PROC_FS */
2453 #ifdef CONFIG_SYSCTL
2456 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2457 void __user *buffer, size_t *lenp, loff_t *ppos)
2459 struct net *net = current->nsproxy->net_ns;
2460 int delay = net->ipv6.sysctl.flush_delay;
2462 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2463 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2469 ctl_table ipv6_route_table_template[] = {
2471 .procname = "flush",
2472 .data = &init_net.ipv6.sysctl.flush_delay,
2473 .maxlen = sizeof(int),
2475 .proc_handler = &ipv6_sysctl_rtcache_flush
2478 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2479 .procname = "gc_thresh",
2480 .data = &ip6_dst_ops_template.gc_thresh,
2481 .maxlen = sizeof(int),
2483 .proc_handler = &proc_dointvec,
2486 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2487 .procname = "max_size",
2488 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2489 .maxlen = sizeof(int),
2491 .proc_handler = &proc_dointvec,
2494 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2495 .procname = "gc_min_interval",
2496 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2497 .maxlen = sizeof(int),
2499 .proc_handler = &proc_dointvec_jiffies,
2500 .strategy = &sysctl_jiffies,
2503 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2504 .procname = "gc_timeout",
2505 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2506 .maxlen = sizeof(int),
2508 .proc_handler = &proc_dointvec_jiffies,
2509 .strategy = &sysctl_jiffies,
2512 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2513 .procname = "gc_interval",
2514 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2515 .maxlen = sizeof(int),
2517 .proc_handler = &proc_dointvec_jiffies,
2518 .strategy = &sysctl_jiffies,
2521 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2522 .procname = "gc_elasticity",
2523 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2524 .maxlen = sizeof(int),
2526 .proc_handler = &proc_dointvec_jiffies,
2527 .strategy = &sysctl_jiffies,
2530 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2531 .procname = "mtu_expires",
2532 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2533 .maxlen = sizeof(int),
2535 .proc_handler = &proc_dointvec_jiffies,
2536 .strategy = &sysctl_jiffies,
2539 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2540 .procname = "min_adv_mss",
2541 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2542 .maxlen = sizeof(int),
2544 .proc_handler = &proc_dointvec_jiffies,
2545 .strategy = &sysctl_jiffies,
2548 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2549 .procname = "gc_min_interval_ms",
2550 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2551 .maxlen = sizeof(int),
2553 .proc_handler = &proc_dointvec_ms_jiffies,
2554 .strategy = &sysctl_ms_jiffies,
2559 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2561 struct ctl_table *table;
2563 table = kmemdup(ipv6_route_table_template,
2564 sizeof(ipv6_route_table_template),
2568 table[0].data = &net->ipv6.sysctl.flush_delay;
2569 table[1].data = &net->ipv6.ip6_dst_ops->gc_thresh;
2570 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2571 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2572 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2573 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2574 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2575 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2576 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2583 static int ip6_route_net_init(struct net *net)
2588 net->ipv6.ip6_dst_ops = kmemdup(&ip6_dst_ops_template,
2589 sizeof(*net->ipv6.ip6_dst_ops),
2591 if (!net->ipv6.ip6_dst_ops)
2593 net->ipv6.ip6_dst_ops->dst_net = net;
2595 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2596 sizeof(*net->ipv6.ip6_null_entry),
2598 if (!net->ipv6.ip6_null_entry)
2599 goto out_ip6_dst_ops;
2600 net->ipv6.ip6_null_entry->u.dst.path =
2601 (struct dst_entry *)net->ipv6.ip6_null_entry;
2602 net->ipv6.ip6_null_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2604 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2605 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2606 sizeof(*net->ipv6.ip6_prohibit_entry),
2608 if (!net->ipv6.ip6_prohibit_entry) {
2609 kfree(net->ipv6.ip6_null_entry);
2612 net->ipv6.ip6_prohibit_entry->u.dst.path =
2613 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2614 net->ipv6.ip6_prohibit_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2616 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2617 sizeof(*net->ipv6.ip6_blk_hole_entry),
2619 if (!net->ipv6.ip6_blk_hole_entry) {
2620 kfree(net->ipv6.ip6_null_entry);
2621 kfree(net->ipv6.ip6_prohibit_entry);
2624 net->ipv6.ip6_blk_hole_entry->u.dst.path =
2625 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2626 net->ipv6.ip6_blk_hole_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2629 #ifdef CONFIG_PROC_FS
2630 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2631 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2633 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2640 kfree(net->ipv6.ip6_dst_ops);
2644 static void ip6_route_net_exit(struct net *net)
2646 #ifdef CONFIG_PROC_FS
2647 proc_net_remove(net, "ipv6_route");
2648 proc_net_remove(net, "rt6_stats");
2650 kfree(net->ipv6.ip6_null_entry);
2651 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2652 kfree(net->ipv6.ip6_prohibit_entry);
2653 kfree(net->ipv6.ip6_blk_hole_entry);
2655 kfree(net->ipv6.ip6_dst_ops);
2658 static struct pernet_operations ip6_route_net_ops = {
2659 .init = ip6_route_net_init,
2660 .exit = ip6_route_net_exit,
2663 static struct notifier_block ip6_route_dev_notifier = {
2664 .notifier_call = ip6_route_dev_notify,
2668 int __init ip6_route_init(void)
2673 ip6_dst_ops_template.kmem_cachep =
2674 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2675 SLAB_HWCACHE_ALIGN, NULL);
2676 if (!ip6_dst_ops_template.kmem_cachep)
2679 ret = register_pernet_subsys(&ip6_route_net_ops);
2681 goto out_kmem_cache;
2683 /* Registering of the loopback is done before this portion of code,
2684 * the loopback reference in rt6_info will not be taken, do it
2685 * manually for init_net */
2686 init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2687 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2688 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2689 init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2690 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2691 init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2692 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2696 goto out_register_subsys;
2702 ret = fib6_rules_init();
2707 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2708 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2709 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2710 goto fib6_rules_init;
2712 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2714 goto fib6_rules_init;
2720 fib6_rules_cleanup();
2725 out_register_subsys:
2726 unregister_pernet_subsys(&ip6_route_net_ops);
2728 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2732 void ip6_route_cleanup(void)
2734 unregister_netdevice_notifier(&ip6_route_dev_notifier);
2735 fib6_rules_cleanup();
2738 unregister_pernet_subsys(&ip6_route_net_ops);
2739 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);