2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/types.h>
30 #include <linux/times.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/route.h>
35 #include <linux/netdevice.h>
36 #include <linux/in6.h>
37 #include <linux/mroute6.h>
38 #include <linux/init.h>
39 #include <linux/if_arp.h>
40 #include <linux/proc_fs.h>
41 #include <linux/seq_file.h>
42 #include <linux/nsproxy.h>
43 #include <linux/slab.h>
44 #include <net/net_namespace.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
58 #include <asm/uaccess.h>
61 #include <linux/sysctl.h>
64 /* Set to 3 to get tracing. */
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
72 #define RT6_TRACE(x...) do { ; } while (0)
75 #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 = cpu_to_be16(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 .entries = ATOMIC_INIT(0),
115 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
119 static struct dst_ops ip6_dst_blackhole_ops = {
121 .protocol = cpu_to_be16(ETH_P_IPV6),
122 .destroy = ip6_dst_destroy,
123 .check = ip6_dst_check,
124 .update_pmtu = ip6_rt_blackhole_update_pmtu,
125 .entries = ATOMIC_INIT(0),
128 static struct rt6_info ip6_null_entry_template = {
130 .__refcnt = ATOMIC_INIT(1),
133 .error = -ENETUNREACH,
134 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
135 .input = ip6_pkt_discard,
136 .output = ip6_pkt_discard_out,
138 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
139 .rt6i_protocol = RTPROT_KERNEL,
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 static struct rt6_info ip6_prohibit_entry_template = {
151 .__refcnt = ATOMIC_INIT(1),
155 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
156 .input = ip6_pkt_prohibit,
157 .output = ip6_pkt_prohibit_out,
159 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
160 .rt6i_protocol = RTPROT_KERNEL,
161 .rt6i_metric = ~(u32) 0,
162 .rt6i_ref = ATOMIC_INIT(1),
165 static struct rt6_info ip6_blk_hole_entry_template = {
167 .__refcnt = ATOMIC_INIT(1),
171 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
172 .input = dst_discard,
173 .output = dst_discard,
175 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
176 .rt6i_protocol = RTPROT_KERNEL,
177 .rt6i_metric = ~(u32) 0,
178 .rt6i_ref = ATOMIC_INIT(1),
183 /* allocate dst with ip6_dst_ops */
184 static inline struct rt6_info *ip6_dst_alloc(struct dst_ops *ops)
186 return (struct rt6_info *)dst_alloc(ops);
189 static void ip6_dst_destroy(struct dst_entry *dst)
191 struct rt6_info *rt = (struct rt6_info *)dst;
192 struct inet6_dev *idev = rt->rt6i_idev;
195 rt->rt6i_idev = NULL;
200 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
203 struct rt6_info *rt = (struct rt6_info *)dst;
204 struct inet6_dev *idev = rt->rt6i_idev;
205 struct net_device *loopback_dev =
206 dev_net(dev)->loopback_dev;
208 if (dev != loopback_dev && idev != NULL && idev->dev == dev) {
209 struct inet6_dev *loopback_idev =
210 in6_dev_get(loopback_dev);
211 if (loopback_idev != NULL) {
212 rt->rt6i_idev = loopback_idev;
218 static __inline__ int rt6_check_expired(const struct rt6_info *rt)
220 return (rt->rt6i_flags & RTF_EXPIRES) &&
221 time_after(jiffies, rt->rt6i_expires);
224 static inline int rt6_need_strict(struct in6_addr *daddr)
226 return ipv6_addr_type(daddr) &
227 (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK);
231 * Route lookup. Any table->tb6_lock is implied.
234 static inline struct rt6_info *rt6_device_match(struct net *net,
236 struct in6_addr *saddr,
240 struct rt6_info *local = NULL;
241 struct rt6_info *sprt;
243 if (!oif && ipv6_addr_any(saddr))
246 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
247 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) {
255 if (flags & RT6_LOOKUP_F_IFACE && oif)
257 if (local && (!oif ||
258 local->rt6i_idev->dev->ifindex == oif))
264 if (ipv6_chk_addr(net, saddr, dev,
265 flags & RT6_LOOKUP_F_IFACE))
274 if (flags & RT6_LOOKUP_F_IFACE)
275 return net->ipv6.ip6_null_entry;
281 #ifdef CONFIG_IPV6_ROUTER_PREF
282 static void rt6_probe(struct rt6_info *rt)
284 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
286 * Okay, this does not seem to be appropriate
287 * for now, however, we need to check if it
288 * is really so; aka Router Reachability Probing.
290 * Router Reachability Probe MUST be rate-limited
291 * to no more than one per minute.
293 if (!neigh || (neigh->nud_state & NUD_VALID))
295 read_lock_bh(&neigh->lock);
296 if (!(neigh->nud_state & NUD_VALID) &&
297 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
298 struct in6_addr mcaddr;
299 struct in6_addr *target;
301 neigh->updated = jiffies;
302 read_unlock_bh(&neigh->lock);
304 target = (struct in6_addr *)&neigh->primary_key;
305 addrconf_addr_solict_mult(target, &mcaddr);
306 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
308 read_unlock_bh(&neigh->lock);
311 static inline void rt6_probe(struct rt6_info *rt)
317 * Default Router Selection (RFC 2461 6.3.6)
319 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
321 struct net_device *dev = rt->rt6i_dev;
322 if (!oif || dev->ifindex == oif)
324 if ((dev->flags & IFF_LOOPBACK) &&
325 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
330 static inline int rt6_check_neigh(struct rt6_info *rt)
332 struct neighbour *neigh = rt->rt6i_nexthop;
334 if (rt->rt6i_flags & RTF_NONEXTHOP ||
335 !(rt->rt6i_flags & RTF_GATEWAY))
338 read_lock_bh(&neigh->lock);
339 if (neigh->nud_state & NUD_VALID)
341 #ifdef CONFIG_IPV6_ROUTER_PREF
342 else if (neigh->nud_state & NUD_FAILED)
347 read_unlock_bh(&neigh->lock);
353 static int rt6_score_route(struct rt6_info *rt, int oif,
358 m = rt6_check_dev(rt, oif);
359 if (!m && (strict & RT6_LOOKUP_F_IFACE))
361 #ifdef CONFIG_IPV6_ROUTER_PREF
362 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
364 n = rt6_check_neigh(rt);
365 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
370 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
371 int *mpri, struct rt6_info *match)
375 if (rt6_check_expired(rt))
378 m = rt6_score_route(rt, oif, strict);
383 if (strict & RT6_LOOKUP_F_REACHABLE)
387 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
395 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
396 struct rt6_info *rr_head,
397 u32 metric, int oif, int strict)
399 struct rt6_info *rt, *match;
403 for (rt = rr_head; rt && rt->rt6i_metric == metric;
404 rt = rt->dst.rt6_next)
405 match = find_match(rt, oif, strict, &mpri, match);
406 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
407 rt = rt->dst.rt6_next)
408 match = find_match(rt, oif, strict, &mpri, match);
413 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
415 struct rt6_info *match, *rt0;
418 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
419 __func__, fn->leaf, oif);
423 fn->rr_ptr = rt0 = fn->leaf;
425 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
428 (strict & RT6_LOOKUP_F_REACHABLE)) {
429 struct rt6_info *next = rt0->dst.rt6_next;
431 /* no entries matched; do round-robin */
432 if (!next || next->rt6i_metric != rt0->rt6i_metric)
439 RT6_TRACE("%s() => %p\n",
442 net = dev_net(rt0->rt6i_dev);
443 return match ? match : net->ipv6.ip6_null_entry;
446 #ifdef CONFIG_IPV6_ROUTE_INFO
447 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
448 struct in6_addr *gwaddr)
450 struct net *net = dev_net(dev);
451 struct route_info *rinfo = (struct route_info *) opt;
452 struct in6_addr prefix_buf, *prefix;
454 unsigned long lifetime;
457 if (len < sizeof(struct route_info)) {
461 /* Sanity check for prefix_len and length */
462 if (rinfo->length > 3) {
464 } else if (rinfo->prefix_len > 128) {
466 } else if (rinfo->prefix_len > 64) {
467 if (rinfo->length < 2) {
470 } else if (rinfo->prefix_len > 0) {
471 if (rinfo->length < 1) {
476 pref = rinfo->route_pref;
477 if (pref == ICMPV6_ROUTER_PREF_INVALID)
480 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
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 (!addrconf_finite_timeout(lifetime)) {
509 rt->rt6i_flags &= ~RTF_EXPIRES;
511 rt->rt6i_expires = jiffies + HZ * lifetime;
512 rt->rt6i_flags |= RTF_EXPIRES;
514 dst_release(&rt->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->fl6_src, fl->oif, flags);
550 BACKTRACK(net, &fl->fl6_src);
552 dst_use(&rt->dst, jiffies);
553 read_unlock_bh(&table->tb6_lock);
558 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
559 const 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 = dev_net(rt->rt6i_dev),
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 struct neighbour *neigh;
628 int attempts = !in_softirq();
630 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
631 if (rt->rt6i_dst.plen != 128 &&
632 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
633 rt->rt6i_flags |= RTF_ANYCAST;
634 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
637 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
638 rt->rt6i_dst.plen = 128;
639 rt->rt6i_flags |= RTF_CACHE;
640 rt->dst.flags |= DST_HOST;
642 #ifdef CONFIG_IPV6_SUBTREES
643 if (rt->rt6i_src.plen && saddr) {
644 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
645 rt->rt6i_src.plen = 128;
650 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
652 struct net *net = dev_net(rt->rt6i_dev);
653 int saved_rt_min_interval =
654 net->ipv6.sysctl.ip6_rt_gc_min_interval;
655 int saved_rt_elasticity =
656 net->ipv6.sysctl.ip6_rt_gc_elasticity;
658 if (attempts-- > 0) {
659 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
660 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
662 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
664 net->ipv6.sysctl.ip6_rt_gc_elasticity =
666 net->ipv6.sysctl.ip6_rt_gc_min_interval =
667 saved_rt_min_interval;
673 "ipv6: Neighbour table overflow.\n");
677 rt->rt6i_nexthop = neigh;
684 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
686 struct rt6_info *rt = ip6_rt_copy(ort);
688 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
689 rt->rt6i_dst.plen = 128;
690 rt->rt6i_flags |= RTF_CACHE;
691 rt->dst.flags |= DST_HOST;
692 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
697 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
698 struct flowi *fl, int flags)
700 struct fib6_node *fn;
701 struct rt6_info *rt, *nrt;
705 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
707 strict |= flags & RT6_LOOKUP_F_IFACE;
710 read_lock_bh(&table->tb6_lock);
713 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
716 rt = rt6_select(fn, oif, strict | reachable);
718 BACKTRACK(net, &fl->fl6_src);
719 if (rt == net->ipv6.ip6_null_entry ||
720 rt->rt6i_flags & RTF_CACHE)
724 read_unlock_bh(&table->tb6_lock);
726 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
727 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
729 #if CLONE_OFFLINK_ROUTE
730 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
736 dst_release(&rt->dst);
737 rt = nrt ? : net->ipv6.ip6_null_entry;
741 err = ip6_ins_rt(nrt);
750 * Race condition! In the gap, when table->tb6_lock was
751 * released someone could insert this route. Relookup.
753 dst_release(&rt->dst);
762 read_unlock_bh(&table->tb6_lock);
764 rt->dst.lastuse = jiffies;
770 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
771 struct flowi *fl, int flags)
773 return ip6_pol_route(net, table, fl->iif, fl, flags);
776 void ip6_route_input(struct sk_buff *skb)
778 struct ipv6hdr *iph = ipv6_hdr(skb);
779 struct net *net = dev_net(skb->dev);
780 int flags = RT6_LOOKUP_F_HAS_SADDR;
782 .iif = skb->dev->ifindex,
787 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
791 .proto = iph->nexthdr,
794 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
795 flags |= RT6_LOOKUP_F_IFACE;
797 skb_dst_set(skb, fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input));
800 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
801 struct flowi *fl, int flags)
803 return ip6_pol_route(net, table, fl->oif, fl, flags);
806 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
811 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl->fl6_dst))
812 flags |= RT6_LOOKUP_F_IFACE;
814 if (!ipv6_addr_any(&fl->fl6_src))
815 flags |= RT6_LOOKUP_F_HAS_SADDR;
817 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
819 return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
822 EXPORT_SYMBOL(ip6_route_output);
824 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
826 struct rt6_info *ort = (struct rt6_info *) *dstp;
827 struct rt6_info *rt = (struct rt6_info *)
828 dst_alloc(&ip6_dst_blackhole_ops);
829 struct dst_entry *new = NULL;
834 atomic_set(&new->__refcnt, 1);
836 new->input = dst_discard;
837 new->output = dst_discard;
839 memcpy(new->metrics, ort->dst.metrics, RTAX_MAX*sizeof(u32));
840 new->dev = ort->dst.dev;
843 rt->rt6i_idev = ort->rt6i_idev;
845 in6_dev_hold(rt->rt6i_idev);
846 rt->rt6i_expires = 0;
848 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
849 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
852 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
853 #ifdef CONFIG_IPV6_SUBTREES
854 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
862 return new ? 0 : -ENOMEM;
864 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
867 * Destination cache support functions
870 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
874 rt = (struct rt6_info *) dst;
876 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
882 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
884 struct rt6_info *rt = (struct rt6_info *) dst;
887 if (rt->rt6i_flags & RTF_CACHE) {
888 if (rt6_check_expired(rt)) {
900 static void ip6_link_failure(struct sk_buff *skb)
904 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
906 rt = (struct rt6_info *) skb_dst(skb);
908 if (rt->rt6i_flags&RTF_CACHE) {
909 dst_set_expires(&rt->dst, 0);
910 rt->rt6i_flags |= RTF_EXPIRES;
911 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
912 rt->rt6i_node->fn_sernum = -1;
916 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
918 struct rt6_info *rt6 = (struct rt6_info*)dst;
920 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
921 rt6->rt6i_flags |= RTF_MODIFIED;
922 if (mtu < IPV6_MIN_MTU) {
924 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
926 dst->metrics[RTAX_MTU-1] = mtu;
927 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
931 static int ipv6_get_mtu(struct net_device *dev);
933 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
935 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
937 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
938 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
941 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
942 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
943 * IPV6_MAXPLEN is also valid and means: "any MSS,
944 * rely only on pmtu discovery"
946 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
951 static struct dst_entry *icmp6_dst_gc_list;
952 static DEFINE_SPINLOCK(icmp6_dst_lock);
954 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
955 struct neighbour *neigh,
956 const struct in6_addr *addr)
959 struct inet6_dev *idev = in6_dev_get(dev);
960 struct net *net = dev_net(dev);
962 if (unlikely(idev == NULL))
965 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
966 if (unlikely(rt == NULL)) {
975 neigh = ndisc_get_neigh(dev, addr);
981 rt->rt6i_idev = idev;
982 rt->rt6i_nexthop = neigh;
983 atomic_set(&rt->dst.__refcnt, 1);
984 rt->dst.metrics[RTAX_HOPLIMIT-1] = 255;
985 rt->dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
986 rt->dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->dst));
987 rt->dst.output = ip6_output;
989 #if 0 /* there's no chance to use these for ndisc */
990 rt->dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
993 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
994 rt->rt6i_dst.plen = 128;
997 spin_lock_bh(&icmp6_dst_lock);
998 rt->dst.next = icmp6_dst_gc_list;
999 icmp6_dst_gc_list = &rt->dst;
1000 spin_unlock_bh(&icmp6_dst_lock);
1002 fib6_force_start_gc(net);
1008 int icmp6_dst_gc(void)
1010 struct dst_entry *dst, *next, **pprev;
1015 spin_lock_bh(&icmp6_dst_lock);
1016 pprev = &icmp6_dst_gc_list;
1018 while ((dst = *pprev) != NULL) {
1019 if (!atomic_read(&dst->__refcnt)) {
1028 spin_unlock_bh(&icmp6_dst_lock);
1033 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1036 struct dst_entry *dst, **pprev;
1038 spin_lock_bh(&icmp6_dst_lock);
1039 pprev = &icmp6_dst_gc_list;
1040 while ((dst = *pprev) != NULL) {
1041 struct rt6_info *rt = (struct rt6_info *) dst;
1042 if (func(rt, arg)) {
1049 spin_unlock_bh(&icmp6_dst_lock);
1052 static int ip6_dst_gc(struct dst_ops *ops)
1054 unsigned long now = jiffies;
1055 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1056 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1057 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1058 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1059 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1060 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1062 if (time_after(rt_last_gc + rt_min_interval, now) &&
1063 atomic_read(&ops->entries) <= rt_max_size)
1066 net->ipv6.ip6_rt_gc_expire++;
1067 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1068 net->ipv6.ip6_rt_last_gc = now;
1069 if (atomic_read(&ops->entries) < ops->gc_thresh)
1070 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1072 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1073 return atomic_read(&ops->entries) > rt_max_size;
1076 /* Clean host part of a prefix. Not necessary in radix tree,
1077 but results in cleaner routing tables.
1079 Remove it only when all the things will work!
1082 static int ipv6_get_mtu(struct net_device *dev)
1084 int mtu = IPV6_MIN_MTU;
1085 struct inet6_dev *idev;
1088 idev = __in6_dev_get(dev);
1090 mtu = idev->cnf.mtu6;
1095 int ip6_dst_hoplimit(struct dst_entry *dst)
1097 int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1099 struct net_device *dev = dst->dev;
1100 struct inet6_dev *idev;
1103 idev = __in6_dev_get(dev);
1105 hoplimit = idev->cnf.hop_limit;
1107 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1117 int ip6_route_add(struct fib6_config *cfg)
1120 struct net *net = cfg->fc_nlinfo.nl_net;
1121 struct rt6_info *rt = NULL;
1122 struct net_device *dev = NULL;
1123 struct inet6_dev *idev = NULL;
1124 struct fib6_table *table;
1127 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1129 #ifndef CONFIG_IPV6_SUBTREES
1130 if (cfg->fc_src_len)
1133 if (cfg->fc_ifindex) {
1135 dev = dev_get_by_index(net, cfg->fc_ifindex);
1138 idev = in6_dev_get(dev);
1143 if (cfg->fc_metric == 0)
1144 cfg->fc_metric = IP6_RT_PRIO_USER;
1146 table = fib6_new_table(net, cfg->fc_table);
1147 if (table == NULL) {
1152 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1159 rt->dst.obsolete = -1;
1160 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1161 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1164 if (cfg->fc_protocol == RTPROT_UNSPEC)
1165 cfg->fc_protocol = RTPROT_BOOT;
1166 rt->rt6i_protocol = cfg->fc_protocol;
1168 addr_type = ipv6_addr_type(&cfg->fc_dst);
1170 if (addr_type & IPV6_ADDR_MULTICAST)
1171 rt->dst.input = ip6_mc_input;
1172 else if (cfg->fc_flags & RTF_LOCAL)
1173 rt->dst.input = ip6_input;
1175 rt->dst.input = ip6_forward;
1177 rt->dst.output = ip6_output;
1179 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1180 rt->rt6i_dst.plen = cfg->fc_dst_len;
1181 if (rt->rt6i_dst.plen == 128)
1182 rt->dst.flags = DST_HOST;
1184 #ifdef CONFIG_IPV6_SUBTREES
1185 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1186 rt->rt6i_src.plen = cfg->fc_src_len;
1189 rt->rt6i_metric = cfg->fc_metric;
1191 /* We cannot add true routes via loopback here,
1192 they would result in kernel looping; promote them to reject routes
1194 if ((cfg->fc_flags & RTF_REJECT) ||
1195 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK)
1196 && !(cfg->fc_flags&RTF_LOCAL))) {
1197 /* hold loopback dev/idev if we haven't done so. */
1198 if (dev != net->loopback_dev) {
1203 dev = net->loopback_dev;
1205 idev = in6_dev_get(dev);
1211 rt->dst.output = ip6_pkt_discard_out;
1212 rt->dst.input = ip6_pkt_discard;
1213 rt->dst.error = -ENETUNREACH;
1214 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1218 if (cfg->fc_flags & RTF_GATEWAY) {
1219 struct in6_addr *gw_addr;
1222 gw_addr = &cfg->fc_gateway;
1223 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1224 gwa_type = ipv6_addr_type(gw_addr);
1226 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1227 struct rt6_info *grt;
1229 /* IPv6 strictly inhibits using not link-local
1230 addresses as nexthop address.
1231 Otherwise, router will not able to send redirects.
1232 It is very good, but in some (rare!) circumstances
1233 (SIT, PtP, NBMA NOARP links) it is handy to allow
1234 some exceptions. --ANK
1237 if (!(gwa_type&IPV6_ADDR_UNICAST))
1240 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1242 err = -EHOSTUNREACH;
1246 if (dev != grt->rt6i_dev) {
1247 dst_release(&grt->dst);
1251 dev = grt->rt6i_dev;
1252 idev = grt->rt6i_idev;
1254 in6_dev_hold(grt->rt6i_idev);
1256 if (!(grt->rt6i_flags&RTF_GATEWAY))
1258 dst_release(&grt->dst);
1264 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1272 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1273 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1274 if (IS_ERR(rt->rt6i_nexthop)) {
1275 err = PTR_ERR(rt->rt6i_nexthop);
1276 rt->rt6i_nexthop = NULL;
1281 rt->rt6i_flags = cfg->fc_flags;
1288 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1289 int type = nla_type(nla);
1292 if (type > RTAX_MAX) {
1297 rt->dst.metrics[type - 1] = nla_get_u32(nla);
1302 if (dst_metric(&rt->dst, RTAX_HOPLIMIT) == 0)
1303 rt->dst.metrics[RTAX_HOPLIMIT-1] = -1;
1304 if (!dst_mtu(&rt->dst))
1305 rt->dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1306 if (!dst_metric(&rt->dst, RTAX_ADVMSS))
1307 rt->dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->dst));
1309 rt->rt6i_idev = idev;
1310 rt->rt6i_table = table;
1312 cfg->fc_nlinfo.nl_net = dev_net(dev);
1314 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1326 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1329 struct fib6_table *table;
1330 struct net *net = dev_net(rt->rt6i_dev);
1332 if (rt == net->ipv6.ip6_null_entry)
1335 table = rt->rt6i_table;
1336 write_lock_bh(&table->tb6_lock);
1338 err = fib6_del(rt, info);
1339 dst_release(&rt->dst);
1341 write_unlock_bh(&table->tb6_lock);
1346 int ip6_del_rt(struct rt6_info *rt)
1348 struct nl_info info = {
1349 .nl_net = dev_net(rt->rt6i_dev),
1351 return __ip6_del_rt(rt, &info);
1354 static int ip6_route_del(struct fib6_config *cfg)
1356 struct fib6_table *table;
1357 struct fib6_node *fn;
1358 struct rt6_info *rt;
1361 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1365 read_lock_bh(&table->tb6_lock);
1367 fn = fib6_locate(&table->tb6_root,
1368 &cfg->fc_dst, cfg->fc_dst_len,
1369 &cfg->fc_src, cfg->fc_src_len);
1372 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1373 if (cfg->fc_ifindex &&
1374 (rt->rt6i_dev == NULL ||
1375 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1377 if (cfg->fc_flags & RTF_GATEWAY &&
1378 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1380 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1383 read_unlock_bh(&table->tb6_lock);
1385 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1388 read_unlock_bh(&table->tb6_lock);
1396 struct ip6rd_flowi {
1398 struct in6_addr gateway;
1401 static struct rt6_info *__ip6_route_redirect(struct net *net,
1402 struct fib6_table *table,
1406 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1407 struct rt6_info *rt;
1408 struct fib6_node *fn;
1411 * Get the "current" route for this destination and
1412 * check if the redirect has come from approriate router.
1414 * RFC 2461 specifies that redirects should only be
1415 * accepted if they come from the nexthop to the target.
1416 * Due to the way the routes are chosen, this notion
1417 * is a bit fuzzy and one might need to check all possible
1421 read_lock_bh(&table->tb6_lock);
1422 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1424 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1426 * Current route is on-link; redirect is always invalid.
1428 * Seems, previous statement is not true. It could
1429 * be node, which looks for us as on-link (f.e. proxy ndisc)
1430 * But then router serving it might decide, that we should
1431 * know truth 8)8) --ANK (980726).
1433 if (rt6_check_expired(rt))
1435 if (!(rt->rt6i_flags & RTF_GATEWAY))
1437 if (fl->oif != rt->rt6i_dev->ifindex)
1439 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1445 rt = net->ipv6.ip6_null_entry;
1446 BACKTRACK(net, &fl->fl6_src);
1450 read_unlock_bh(&table->tb6_lock);
1455 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1456 struct in6_addr *src,
1457 struct in6_addr *gateway,
1458 struct net_device *dev)
1460 int flags = RT6_LOOKUP_F_HAS_SADDR;
1461 struct net *net = dev_net(dev);
1462 struct ip6rd_flowi rdfl = {
1464 .oif = dev->ifindex,
1474 ipv6_addr_copy(&rdfl.gateway, gateway);
1476 if (rt6_need_strict(dest))
1477 flags |= RT6_LOOKUP_F_IFACE;
1479 return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1480 flags, __ip6_route_redirect);
1483 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1484 struct in6_addr *saddr,
1485 struct neighbour *neigh, u8 *lladdr, int on_link)
1487 struct rt6_info *rt, *nrt = NULL;
1488 struct netevent_redirect netevent;
1489 struct net *net = dev_net(neigh->dev);
1491 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1493 if (rt == net->ipv6.ip6_null_entry) {
1494 if (net_ratelimit())
1495 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1496 "for redirect target\n");
1501 * We have finally decided to accept it.
1504 neigh_update(neigh, lladdr, NUD_STALE,
1505 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1506 NEIGH_UPDATE_F_OVERRIDE|
1507 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1508 NEIGH_UPDATE_F_ISROUTER))
1512 * Redirect received -> path was valid.
1513 * Look, redirects are sent only in response to data packets,
1514 * so that this nexthop apparently is reachable. --ANK
1516 dst_confirm(&rt->dst);
1518 /* Duplicate redirect: silently ignore. */
1519 if (neigh == rt->dst.neighbour)
1522 nrt = ip6_rt_copy(rt);
1526 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1528 nrt->rt6i_flags &= ~RTF_GATEWAY;
1530 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1531 nrt->rt6i_dst.plen = 128;
1532 nrt->dst.flags |= DST_HOST;
1534 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1535 nrt->rt6i_nexthop = neigh_clone(neigh);
1536 /* Reset pmtu, it may be better */
1537 nrt->dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1538 nrt->dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1539 dst_mtu(&nrt->dst));
1541 if (ip6_ins_rt(nrt))
1544 netevent.old = &rt->dst;
1545 netevent.new = &nrt->dst;
1546 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1548 if (rt->rt6i_flags&RTF_CACHE) {
1554 dst_release(&rt->dst);
1558 * Handle ICMP "packet too big" messages
1559 * i.e. Path MTU discovery
1562 static void rt6_do_pmtu_disc(struct in6_addr *daddr, struct in6_addr *saddr,
1563 struct net *net, u32 pmtu, int ifindex)
1565 struct rt6_info *rt, *nrt;
1568 rt = rt6_lookup(net, daddr, saddr, ifindex, 0);
1572 if (pmtu >= dst_mtu(&rt->dst))
1575 if (pmtu < IPV6_MIN_MTU) {
1577 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1578 * MTU (1280) and a fragment header should always be included
1579 * after a node receiving Too Big message reporting PMTU is
1580 * less than the IPv6 Minimum Link MTU.
1582 pmtu = IPV6_MIN_MTU;
1586 /* New mtu received -> path was valid.
1587 They are sent only in response to data packets,
1588 so that this nexthop apparently is reachable. --ANK
1590 dst_confirm(&rt->dst);
1592 /* Host route. If it is static, it would be better
1593 not to override it, but add new one, so that
1594 when cache entry will expire old pmtu
1595 would return automatically.
1597 if (rt->rt6i_flags & RTF_CACHE) {
1598 rt->dst.metrics[RTAX_MTU-1] = pmtu;
1600 rt->dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1601 dst_set_expires(&rt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1602 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1607 Two cases are possible:
1608 1. It is connected route. Action: COW
1609 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1611 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1612 nrt = rt6_alloc_cow(rt, daddr, saddr);
1614 nrt = rt6_alloc_clone(rt, daddr);
1617 nrt->dst.metrics[RTAX_MTU-1] = pmtu;
1619 nrt->dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1621 /* According to RFC 1981, detecting PMTU increase shouldn't be
1622 * happened within 5 mins, the recommended timer is 10 mins.
1623 * Here this route expiration time is set to ip6_rt_mtu_expires
1624 * which is 10 mins. After 10 mins the decreased pmtu is expired
1625 * and detecting PMTU increase will be automatically happened.
1627 dst_set_expires(&nrt->dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1628 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1633 dst_release(&rt->dst);
1636 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1637 struct net_device *dev, u32 pmtu)
1639 struct net *net = dev_net(dev);
1642 * RFC 1981 states that a node "MUST reduce the size of the packets it
1643 * is sending along the path" that caused the Packet Too Big message.
1644 * Since it's not possible in the general case to determine which
1645 * interface was used to send the original packet, we update the MTU
1646 * on the interface that will be used to send future packets. We also
1647 * update the MTU on the interface that received the Packet Too Big in
1648 * case the original packet was forced out that interface with
1649 * SO_BINDTODEVICE or similar. This is the next best thing to the
1650 * correct behaviour, which would be to update the MTU on all
1653 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, 0);
1654 rt6_do_pmtu_disc(daddr, saddr, net, pmtu, dev->ifindex);
1658 * Misc support functions
1661 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1663 struct net *net = dev_net(ort->rt6i_dev);
1664 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1667 rt->dst.input = ort->dst.input;
1668 rt->dst.output = ort->dst.output;
1670 memcpy(rt->dst.metrics, ort->dst.metrics, RTAX_MAX*sizeof(u32));
1671 rt->dst.error = ort->dst.error;
1672 rt->dst.dev = ort->dst.dev;
1674 dev_hold(rt->dst.dev);
1675 rt->rt6i_idev = ort->rt6i_idev;
1677 in6_dev_hold(rt->rt6i_idev);
1678 rt->dst.lastuse = jiffies;
1679 rt->rt6i_expires = 0;
1681 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1682 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1683 rt->rt6i_metric = 0;
1685 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1686 #ifdef CONFIG_IPV6_SUBTREES
1687 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1689 rt->rt6i_table = ort->rt6i_table;
1694 #ifdef CONFIG_IPV6_ROUTE_INFO
1695 static struct rt6_info *rt6_get_route_info(struct net *net,
1696 struct in6_addr *prefix, int prefixlen,
1697 struct in6_addr *gwaddr, int ifindex)
1699 struct fib6_node *fn;
1700 struct rt6_info *rt = NULL;
1701 struct fib6_table *table;
1703 table = fib6_get_table(net, RT6_TABLE_INFO);
1707 write_lock_bh(&table->tb6_lock);
1708 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1712 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1713 if (rt->rt6i_dev->ifindex != ifindex)
1715 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1717 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1723 write_unlock_bh(&table->tb6_lock);
1727 static struct rt6_info *rt6_add_route_info(struct net *net,
1728 struct in6_addr *prefix, int prefixlen,
1729 struct in6_addr *gwaddr, int ifindex,
1732 struct fib6_config cfg = {
1733 .fc_table = RT6_TABLE_INFO,
1734 .fc_metric = IP6_RT_PRIO_USER,
1735 .fc_ifindex = ifindex,
1736 .fc_dst_len = prefixlen,
1737 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1738 RTF_UP | RTF_PREF(pref),
1740 .fc_nlinfo.nlh = NULL,
1741 .fc_nlinfo.nl_net = net,
1744 ipv6_addr_copy(&cfg.fc_dst, prefix);
1745 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1747 /* We should treat it as a default route if prefix length is 0. */
1749 cfg.fc_flags |= RTF_DEFAULT;
1751 ip6_route_add(&cfg);
1753 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1757 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1759 struct rt6_info *rt;
1760 struct fib6_table *table;
1762 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1766 write_lock_bh(&table->tb6_lock);
1767 for (rt = table->tb6_root.leaf; rt; rt=rt->dst.rt6_next) {
1768 if (dev == rt->rt6i_dev &&
1769 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1770 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1775 write_unlock_bh(&table->tb6_lock);
1779 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1780 struct net_device *dev,
1783 struct fib6_config cfg = {
1784 .fc_table = RT6_TABLE_DFLT,
1785 .fc_metric = IP6_RT_PRIO_USER,
1786 .fc_ifindex = dev->ifindex,
1787 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1788 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1790 .fc_nlinfo.nlh = NULL,
1791 .fc_nlinfo.nl_net = dev_net(dev),
1794 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1796 ip6_route_add(&cfg);
1798 return rt6_get_dflt_router(gwaddr, dev);
1801 void rt6_purge_dflt_routers(struct net *net)
1803 struct rt6_info *rt;
1804 struct fib6_table *table;
1806 /* NOTE: Keep consistent with rt6_get_dflt_router */
1807 table = fib6_get_table(net, RT6_TABLE_DFLT);
1812 read_lock_bh(&table->tb6_lock);
1813 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
1814 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1816 read_unlock_bh(&table->tb6_lock);
1821 read_unlock_bh(&table->tb6_lock);
1824 static void rtmsg_to_fib6_config(struct net *net,
1825 struct in6_rtmsg *rtmsg,
1826 struct fib6_config *cfg)
1828 memset(cfg, 0, sizeof(*cfg));
1830 cfg->fc_table = RT6_TABLE_MAIN;
1831 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1832 cfg->fc_metric = rtmsg->rtmsg_metric;
1833 cfg->fc_expires = rtmsg->rtmsg_info;
1834 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1835 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1836 cfg->fc_flags = rtmsg->rtmsg_flags;
1838 cfg->fc_nlinfo.nl_net = net;
1840 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1841 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1842 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1845 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1847 struct fib6_config cfg;
1848 struct in6_rtmsg rtmsg;
1852 case SIOCADDRT: /* Add a route */
1853 case SIOCDELRT: /* Delete a route */
1854 if (!capable(CAP_NET_ADMIN))
1856 err = copy_from_user(&rtmsg, arg,
1857 sizeof(struct in6_rtmsg));
1861 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1866 err = ip6_route_add(&cfg);
1869 err = ip6_route_del(&cfg);
1883 * Drop the packet on the floor
1886 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1889 struct dst_entry *dst = skb_dst(skb);
1890 switch (ipstats_mib_noroutes) {
1891 case IPSTATS_MIB_INNOROUTES:
1892 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1893 if (type == IPV6_ADDR_ANY) {
1894 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1895 IPSTATS_MIB_INADDRERRORS);
1899 case IPSTATS_MIB_OUTNOROUTES:
1900 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1901 ipstats_mib_noroutes);
1904 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1909 static int ip6_pkt_discard(struct sk_buff *skb)
1911 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1914 static int ip6_pkt_discard_out(struct sk_buff *skb)
1916 skb->dev = skb_dst(skb)->dev;
1917 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1920 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1922 static int ip6_pkt_prohibit(struct sk_buff *skb)
1924 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1927 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1929 skb->dev = skb_dst(skb)->dev;
1930 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1936 * Allocate a dst for local (unicast / anycast) address.
1939 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1940 const struct in6_addr *addr,
1943 struct net *net = dev_net(idev->dev);
1944 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1945 struct neighbour *neigh;
1948 return ERR_PTR(-ENOMEM);
1950 dev_hold(net->loopback_dev);
1953 rt->dst.flags = DST_HOST;
1954 rt->dst.input = ip6_input;
1955 rt->dst.output = ip6_output;
1956 rt->rt6i_dev = net->loopback_dev;
1957 rt->rt6i_idev = idev;
1958 rt->dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1959 rt->dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->dst));
1960 rt->dst.metrics[RTAX_HOPLIMIT-1] = -1;
1961 rt->dst.obsolete = -1;
1963 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1965 rt->rt6i_flags |= RTF_ANYCAST;
1967 rt->rt6i_flags |= RTF_LOCAL;
1968 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1969 if (IS_ERR(neigh)) {
1972 /* We are casting this because that is the return
1973 * value type. But an errno encoded pointer is the
1974 * same regardless of the underlying pointer type,
1975 * and that's what we are returning. So this is OK.
1977 return (struct rt6_info *) neigh;
1979 rt->rt6i_nexthop = neigh;
1981 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1982 rt->rt6i_dst.plen = 128;
1983 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1985 atomic_set(&rt->dst.__refcnt, 1);
1990 struct arg_dev_net {
1991 struct net_device *dev;
1995 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1997 struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1998 struct net *net = ((struct arg_dev_net *)arg)->net;
2000 if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
2001 rt != net->ipv6.ip6_null_entry) {
2002 RT6_TRACE("deleted by ifdown %p\n", rt);
2008 void rt6_ifdown(struct net *net, struct net_device *dev)
2010 struct arg_dev_net adn = {
2015 fib6_clean_all(net, fib6_ifdown, 0, &adn);
2016 icmp6_clean_all(fib6_ifdown, &adn);
2019 struct rt6_mtu_change_arg
2021 struct net_device *dev;
2025 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2027 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2028 struct inet6_dev *idev;
2029 struct net *net = dev_net(arg->dev);
2031 /* In IPv6 pmtu discovery is not optional,
2032 so that RTAX_MTU lock cannot disable it.
2033 We still use this lock to block changes
2034 caused by addrconf/ndisc.
2037 idev = __in6_dev_get(arg->dev);
2041 /* For administrative MTU increase, there is no way to discover
2042 IPv6 PMTU increase, so PMTU increase should be updated here.
2043 Since RFC 1981 doesn't include administrative MTU increase
2044 update PMTU increase is a MUST. (i.e. jumbo frame)
2047 If new MTU is less than route PMTU, this new MTU will be the
2048 lowest MTU in the path, update the route PMTU to reflect PMTU
2049 decreases; if new MTU is greater than route PMTU, and the
2050 old MTU is the lowest MTU in the path, update the route PMTU
2051 to reflect the increase. In this case if the other nodes' MTU
2052 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2055 if (rt->rt6i_dev == arg->dev &&
2056 !dst_metric_locked(&rt->dst, RTAX_MTU) &&
2057 (dst_mtu(&rt->dst) >= arg->mtu ||
2058 (dst_mtu(&rt->dst) < arg->mtu &&
2059 dst_mtu(&rt->dst) == idev->cnf.mtu6))) {
2060 rt->dst.metrics[RTAX_MTU-1] = arg->mtu;
2061 rt->dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
2066 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2068 struct rt6_mtu_change_arg arg = {
2073 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2076 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2077 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2078 [RTA_OIF] = { .type = NLA_U32 },
2079 [RTA_IIF] = { .type = NLA_U32 },
2080 [RTA_PRIORITY] = { .type = NLA_U32 },
2081 [RTA_METRICS] = { .type = NLA_NESTED },
2084 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2085 struct fib6_config *cfg)
2088 struct nlattr *tb[RTA_MAX+1];
2091 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2096 rtm = nlmsg_data(nlh);
2097 memset(cfg, 0, sizeof(*cfg));
2099 cfg->fc_table = rtm->rtm_table;
2100 cfg->fc_dst_len = rtm->rtm_dst_len;
2101 cfg->fc_src_len = rtm->rtm_src_len;
2102 cfg->fc_flags = RTF_UP;
2103 cfg->fc_protocol = rtm->rtm_protocol;
2105 if (rtm->rtm_type == RTN_UNREACHABLE)
2106 cfg->fc_flags |= RTF_REJECT;
2108 if (rtm->rtm_type == RTN_LOCAL)
2109 cfg->fc_flags |= RTF_LOCAL;
2111 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2112 cfg->fc_nlinfo.nlh = nlh;
2113 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2115 if (tb[RTA_GATEWAY]) {
2116 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2117 cfg->fc_flags |= RTF_GATEWAY;
2121 int plen = (rtm->rtm_dst_len + 7) >> 3;
2123 if (nla_len(tb[RTA_DST]) < plen)
2126 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2130 int plen = (rtm->rtm_src_len + 7) >> 3;
2132 if (nla_len(tb[RTA_SRC]) < plen)
2135 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2139 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2141 if (tb[RTA_PRIORITY])
2142 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2144 if (tb[RTA_METRICS]) {
2145 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2146 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2150 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2157 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2159 struct fib6_config cfg;
2162 err = rtm_to_fib6_config(skb, nlh, &cfg);
2166 return ip6_route_del(&cfg);
2169 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2171 struct fib6_config cfg;
2174 err = rtm_to_fib6_config(skb, nlh, &cfg);
2178 return ip6_route_add(&cfg);
2181 static inline size_t rt6_nlmsg_size(void)
2183 return NLMSG_ALIGN(sizeof(struct rtmsg))
2184 + nla_total_size(16) /* RTA_SRC */
2185 + nla_total_size(16) /* RTA_DST */
2186 + nla_total_size(16) /* RTA_GATEWAY */
2187 + nla_total_size(16) /* RTA_PREFSRC */
2188 + nla_total_size(4) /* RTA_TABLE */
2189 + nla_total_size(4) /* RTA_IIF */
2190 + nla_total_size(4) /* RTA_OIF */
2191 + nla_total_size(4) /* RTA_PRIORITY */
2192 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2193 + nla_total_size(sizeof(struct rta_cacheinfo));
2196 static int rt6_fill_node(struct net *net,
2197 struct sk_buff *skb, struct rt6_info *rt,
2198 struct in6_addr *dst, struct in6_addr *src,
2199 int iif, int type, u32 pid, u32 seq,
2200 int prefix, int nowait, unsigned int flags)
2203 struct nlmsghdr *nlh;
2207 if (prefix) { /* user wants prefix routes only */
2208 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2209 /* success since this is not a prefix route */
2214 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2218 rtm = nlmsg_data(nlh);
2219 rtm->rtm_family = AF_INET6;
2220 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2221 rtm->rtm_src_len = rt->rt6i_src.plen;
2224 table = rt->rt6i_table->tb6_id;
2226 table = RT6_TABLE_UNSPEC;
2227 rtm->rtm_table = table;
2228 NLA_PUT_U32(skb, RTA_TABLE, table);
2229 if (rt->rt6i_flags&RTF_REJECT)
2230 rtm->rtm_type = RTN_UNREACHABLE;
2231 else if (rt->rt6i_flags&RTF_LOCAL)
2232 rtm->rtm_type = RTN_LOCAL;
2233 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2234 rtm->rtm_type = RTN_LOCAL;
2236 rtm->rtm_type = RTN_UNICAST;
2238 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2239 rtm->rtm_protocol = rt->rt6i_protocol;
2240 if (rt->rt6i_flags&RTF_DYNAMIC)
2241 rtm->rtm_protocol = RTPROT_REDIRECT;
2242 else if (rt->rt6i_flags & RTF_ADDRCONF)
2243 rtm->rtm_protocol = RTPROT_KERNEL;
2244 else if (rt->rt6i_flags&RTF_DEFAULT)
2245 rtm->rtm_protocol = RTPROT_RA;
2247 if (rt->rt6i_flags&RTF_CACHE)
2248 rtm->rtm_flags |= RTM_F_CLONED;
2251 NLA_PUT(skb, RTA_DST, 16, dst);
2252 rtm->rtm_dst_len = 128;
2253 } else if (rtm->rtm_dst_len)
2254 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2255 #ifdef CONFIG_IPV6_SUBTREES
2257 NLA_PUT(skb, RTA_SRC, 16, src);
2258 rtm->rtm_src_len = 128;
2259 } else if (rtm->rtm_src_len)
2260 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2263 #ifdef CONFIG_IPV6_MROUTE
2264 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2265 int err = ip6mr_get_route(net, skb, rtm, nowait);
2270 goto nla_put_failure;
2272 if (err == -EMSGSIZE)
2273 goto nla_put_failure;
2278 NLA_PUT_U32(skb, RTA_IIF, iif);
2280 struct inet6_dev *idev = ip6_dst_idev(&rt->dst);
2281 struct in6_addr saddr_buf;
2282 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2283 dst, 0, &saddr_buf) == 0)
2284 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2287 if (rtnetlink_put_metrics(skb, rt->dst.metrics) < 0)
2288 goto nla_put_failure;
2290 if (rt->dst.neighbour)
2291 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->dst.neighbour->primary_key);
2294 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2296 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2298 if (!(rt->rt6i_flags & RTF_EXPIRES))
2300 else if (rt->rt6i_expires - jiffies < INT_MAX)
2301 expires = rt->rt6i_expires - jiffies;
2305 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, 0, 0,
2306 expires, rt->dst.error) < 0)
2307 goto nla_put_failure;
2309 return nlmsg_end(skb, nlh);
2312 nlmsg_cancel(skb, nlh);
2316 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2318 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2321 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2322 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2323 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2327 return rt6_fill_node(arg->net,
2328 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2329 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2330 prefix, 0, NLM_F_MULTI);
2333 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2335 struct net *net = sock_net(in_skb->sk);
2336 struct nlattr *tb[RTA_MAX+1];
2337 struct rt6_info *rt;
2338 struct sk_buff *skb;
2343 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2348 memset(&fl, 0, sizeof(fl));
2351 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2354 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2358 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2361 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2365 iif = nla_get_u32(tb[RTA_IIF]);
2368 fl.oif = nla_get_u32(tb[RTA_OIF]);
2371 struct net_device *dev;
2372 dev = __dev_get_by_index(net, iif);
2379 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2385 /* Reserve room for dummy headers, this skb can pass
2386 through good chunk of routing engine.
2388 skb_reset_mac_header(skb);
2389 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2391 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2392 skb_dst_set(skb, &rt->dst);
2394 err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2395 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2396 nlh->nlmsg_seq, 0, 0, 0);
2402 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2407 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2409 struct sk_buff *skb;
2410 struct net *net = info->nl_net;
2415 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2417 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2421 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2422 event, info->pid, seq, 0, 0, 0);
2424 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2425 WARN_ON(err == -EMSGSIZE);
2429 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2430 info->nlh, gfp_any());
2434 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2437 static int ip6_route_dev_notify(struct notifier_block *this,
2438 unsigned long event, void *data)
2440 struct net_device *dev = (struct net_device *)data;
2441 struct net *net = dev_net(dev);
2443 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2444 net->ipv6.ip6_null_entry->dst.dev = dev;
2445 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2446 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2447 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
2448 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2449 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
2450 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2461 #ifdef CONFIG_PROC_FS
2463 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2474 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2476 struct seq_file *m = p_arg;
2478 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2480 #ifdef CONFIG_IPV6_SUBTREES
2481 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2483 seq_puts(m, "00000000000000000000000000000000 00 ");
2486 if (rt->rt6i_nexthop) {
2487 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2489 seq_puts(m, "00000000000000000000000000000000");
2491 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2492 rt->rt6i_metric, atomic_read(&rt->dst.__refcnt),
2493 rt->dst.__use, rt->rt6i_flags,
2494 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2498 static int ipv6_route_show(struct seq_file *m, void *v)
2500 struct net *net = (struct net *)m->private;
2501 fib6_clean_all(net, rt6_info_route, 0, m);
2505 static int ipv6_route_open(struct inode *inode, struct file *file)
2507 return single_open_net(inode, file, ipv6_route_show);
2510 static const struct file_operations ipv6_route_proc_fops = {
2511 .owner = THIS_MODULE,
2512 .open = ipv6_route_open,
2514 .llseek = seq_lseek,
2515 .release = single_release_net,
2518 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2520 struct net *net = (struct net *)seq->private;
2521 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2522 net->ipv6.rt6_stats->fib_nodes,
2523 net->ipv6.rt6_stats->fib_route_nodes,
2524 net->ipv6.rt6_stats->fib_rt_alloc,
2525 net->ipv6.rt6_stats->fib_rt_entries,
2526 net->ipv6.rt6_stats->fib_rt_cache,
2527 atomic_read(&net->ipv6.ip6_dst_ops.entries),
2528 net->ipv6.rt6_stats->fib_discarded_routes);
2533 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2535 return single_open_net(inode, file, rt6_stats_seq_show);
2538 static const struct file_operations rt6_stats_seq_fops = {
2539 .owner = THIS_MODULE,
2540 .open = rt6_stats_seq_open,
2542 .llseek = seq_lseek,
2543 .release = single_release_net,
2545 #endif /* CONFIG_PROC_FS */
2547 #ifdef CONFIG_SYSCTL
2550 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2551 void __user *buffer, size_t *lenp, loff_t *ppos)
2553 struct net *net = current->nsproxy->net_ns;
2554 int delay = net->ipv6.sysctl.flush_delay;
2556 proc_dointvec(ctl, write, buffer, lenp, ppos);
2557 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2563 ctl_table ipv6_route_table_template[] = {
2565 .procname = "flush",
2566 .data = &init_net.ipv6.sysctl.flush_delay,
2567 .maxlen = sizeof(int),
2569 .proc_handler = ipv6_sysctl_rtcache_flush
2572 .procname = "gc_thresh",
2573 .data = &ip6_dst_ops_template.gc_thresh,
2574 .maxlen = sizeof(int),
2576 .proc_handler = proc_dointvec,
2579 .procname = "max_size",
2580 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2581 .maxlen = sizeof(int),
2583 .proc_handler = proc_dointvec,
2586 .procname = "gc_min_interval",
2587 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2588 .maxlen = sizeof(int),
2590 .proc_handler = proc_dointvec_jiffies,
2593 .procname = "gc_timeout",
2594 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2595 .maxlen = sizeof(int),
2597 .proc_handler = proc_dointvec_jiffies,
2600 .procname = "gc_interval",
2601 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2602 .maxlen = sizeof(int),
2604 .proc_handler = proc_dointvec_jiffies,
2607 .procname = "gc_elasticity",
2608 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2609 .maxlen = sizeof(int),
2611 .proc_handler = proc_dointvec,
2614 .procname = "mtu_expires",
2615 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2616 .maxlen = sizeof(int),
2618 .proc_handler = proc_dointvec_jiffies,
2621 .procname = "min_adv_mss",
2622 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2623 .maxlen = sizeof(int),
2625 .proc_handler = proc_dointvec,
2628 .procname = "gc_min_interval_ms",
2629 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2630 .maxlen = sizeof(int),
2632 .proc_handler = proc_dointvec_ms_jiffies,
2637 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2639 struct ctl_table *table;
2641 table = kmemdup(ipv6_route_table_template,
2642 sizeof(ipv6_route_table_template),
2646 table[0].data = &net->ipv6.sysctl.flush_delay;
2647 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2648 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2649 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2650 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2651 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2652 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2653 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2654 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2655 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2662 static int __net_init ip6_route_net_init(struct net *net)
2666 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2667 sizeof(net->ipv6.ip6_dst_ops));
2669 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2670 sizeof(*net->ipv6.ip6_null_entry),
2672 if (!net->ipv6.ip6_null_entry)
2673 goto out_ip6_dst_ops;
2674 net->ipv6.ip6_null_entry->dst.path =
2675 (struct dst_entry *)net->ipv6.ip6_null_entry;
2676 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2678 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2679 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2680 sizeof(*net->ipv6.ip6_prohibit_entry),
2682 if (!net->ipv6.ip6_prohibit_entry)
2683 goto out_ip6_null_entry;
2684 net->ipv6.ip6_prohibit_entry->dst.path =
2685 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2686 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2688 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2689 sizeof(*net->ipv6.ip6_blk_hole_entry),
2691 if (!net->ipv6.ip6_blk_hole_entry)
2692 goto out_ip6_prohibit_entry;
2693 net->ipv6.ip6_blk_hole_entry->dst.path =
2694 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2695 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
2698 net->ipv6.sysctl.flush_delay = 0;
2699 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2700 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2701 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2702 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2703 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2704 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2705 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2707 #ifdef CONFIG_PROC_FS
2708 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2709 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2711 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2717 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2718 out_ip6_prohibit_entry:
2719 kfree(net->ipv6.ip6_prohibit_entry);
2721 kfree(net->ipv6.ip6_null_entry);
2727 static void __net_exit ip6_route_net_exit(struct net *net)
2729 #ifdef CONFIG_PROC_FS
2730 proc_net_remove(net, "ipv6_route");
2731 proc_net_remove(net, "rt6_stats");
2733 kfree(net->ipv6.ip6_null_entry);
2734 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2735 kfree(net->ipv6.ip6_prohibit_entry);
2736 kfree(net->ipv6.ip6_blk_hole_entry);
2740 static struct pernet_operations ip6_route_net_ops = {
2741 .init = ip6_route_net_init,
2742 .exit = ip6_route_net_exit,
2745 static struct notifier_block ip6_route_dev_notifier = {
2746 .notifier_call = ip6_route_dev_notify,
2750 int __init ip6_route_init(void)
2755 ip6_dst_ops_template.kmem_cachep =
2756 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2757 SLAB_HWCACHE_ALIGN, NULL);
2758 if (!ip6_dst_ops_template.kmem_cachep)
2761 ret = register_pernet_subsys(&ip6_route_net_ops);
2763 goto out_kmem_cache;
2765 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2767 /* Registering of the loopback is done before this portion of code,
2768 * the loopback reference in rt6_info will not be taken, do it
2769 * manually for init_net */
2770 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
2771 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2772 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2773 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
2774 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2775 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
2776 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2780 goto out_register_subsys;
2786 ret = fib6_rules_init();
2791 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2792 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2793 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2794 goto fib6_rules_init;
2796 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2798 goto fib6_rules_init;
2804 fib6_rules_cleanup();
2809 out_register_subsys:
2810 unregister_pernet_subsys(&ip6_route_net_ops);
2812 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2816 void ip6_route_cleanup(void)
2818 unregister_netdevice_notifier(&ip6_route_dev_notifier);
2819 fib6_rules_cleanup();
2822 unregister_pernet_subsys(&ip6_route_net_ops);
2823 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);