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 <net/net_namespace.h>
46 #include <net/ip6_fib.h>
47 #include <net/ip6_route.h>
48 #include <net/ndisc.h>
49 #include <net/addrconf.h>
51 #include <linux/rtnetlink.h>
54 #include <net/netevent.h>
55 #include <net/netlink.h>
57 #include <asm/uaccess.h>
60 #include <linux/sysctl.h>
63 /* Set to 3 to get tracing. */
67 #define RDBG(x) printk x
68 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
71 #define RT6_TRACE(x...) do { ; } while (0)
74 #define CLONE_OFFLINK_ROUTE 0
76 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort);
77 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
78 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
79 static void ip6_dst_destroy(struct dst_entry *);
80 static void ip6_dst_ifdown(struct dst_entry *,
81 struct net_device *dev, int how);
82 static int ip6_dst_gc(struct dst_ops *ops);
84 static int ip6_pkt_discard(struct sk_buff *skb);
85 static int ip6_pkt_discard_out(struct sk_buff *skb);
86 static void ip6_link_failure(struct sk_buff *skb);
87 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu);
89 #ifdef CONFIG_IPV6_ROUTE_INFO
90 static struct rt6_info *rt6_add_route_info(struct net *net,
91 struct in6_addr *prefix, int prefixlen,
92 struct in6_addr *gwaddr, int ifindex,
94 static struct rt6_info *rt6_get_route_info(struct net *net,
95 struct in6_addr *prefix, int prefixlen,
96 struct in6_addr *gwaddr, int ifindex);
99 static struct dst_ops ip6_dst_ops_template = {
101 .protocol = cpu_to_be16(ETH_P_IPV6),
104 .check = ip6_dst_check,
105 .destroy = ip6_dst_destroy,
106 .ifdown = ip6_dst_ifdown,
107 .negative_advice = ip6_negative_advice,
108 .link_failure = ip6_link_failure,
109 .update_pmtu = ip6_rt_update_pmtu,
110 .local_out = __ip6_local_out,
111 .entries = ATOMIC_INIT(0),
114 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, u32 mtu)
118 static struct dst_ops ip6_dst_blackhole_ops = {
120 .protocol = cpu_to_be16(ETH_P_IPV6),
121 .destroy = ip6_dst_destroy,
122 .check = ip6_dst_check,
123 .update_pmtu = ip6_rt_blackhole_update_pmtu,
124 .entries = ATOMIC_INIT(0),
127 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,
139 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
140 .rt6i_protocol = RTPROT_KERNEL,
141 .rt6i_metric = ~(u32) 0,
142 .rt6i_ref = ATOMIC_INIT(1),
145 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
147 static int ip6_pkt_prohibit(struct sk_buff *skb);
148 static int ip6_pkt_prohibit_out(struct sk_buff *skb);
150 static struct rt6_info ip6_prohibit_entry_template = {
153 .__refcnt = ATOMIC_INIT(1),
157 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
158 .input = ip6_pkt_prohibit,
159 .output = ip6_pkt_prohibit_out,
162 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
163 .rt6i_protocol = RTPROT_KERNEL,
164 .rt6i_metric = ~(u32) 0,
165 .rt6i_ref = ATOMIC_INIT(1),
168 static struct rt6_info ip6_blk_hole_entry_template = {
171 .__refcnt = ATOMIC_INIT(1),
175 .metrics = { [RTAX_HOPLIMIT - 1] = 255, },
176 .input = dst_discard,
177 .output = dst_discard,
180 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
181 .rt6i_protocol = RTPROT_KERNEL,
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_net(dev)->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 | IPV6_ADDR_LOOPBACK));
236 * Route lookup. Any table->tb6_lock is implied.
239 static inline struct rt6_info *rt6_device_match(struct net *net,
241 struct in6_addr *saddr,
245 struct rt6_info *local = NULL;
246 struct rt6_info *sprt;
248 if (!oif && ipv6_addr_any(saddr))
251 for (sprt = rt; sprt; sprt = sprt->u.dst.rt6_next) {
252 struct net_device *dev = sprt->rt6i_dev;
255 if (dev->ifindex == oif)
257 if (dev->flags & IFF_LOOPBACK) {
258 if (sprt->rt6i_idev == NULL ||
259 sprt->rt6i_idev->dev->ifindex != oif) {
260 if (flags & RT6_LOOKUP_F_IFACE && oif)
262 if (local && (!oif ||
263 local->rt6i_idev->dev->ifindex == oif))
269 if (ipv6_chk_addr(net, saddr, dev,
270 flags & RT6_LOOKUP_F_IFACE))
279 if (flags & RT6_LOOKUP_F_IFACE)
280 return net->ipv6.ip6_null_entry;
286 #ifdef CONFIG_IPV6_ROUTER_PREF
287 static void rt6_probe(struct rt6_info *rt)
289 struct neighbour *neigh = rt ? rt->rt6i_nexthop : NULL;
291 * Okay, this does not seem to be appropriate
292 * for now, however, we need to check if it
293 * is really so; aka Router Reachability Probing.
295 * Router Reachability Probe MUST be rate-limited
296 * to no more than one per minute.
298 if (!neigh || (neigh->nud_state & NUD_VALID))
300 read_lock_bh(&neigh->lock);
301 if (!(neigh->nud_state & NUD_VALID) &&
302 time_after(jiffies, neigh->updated + rt->rt6i_idev->cnf.rtr_probe_interval)) {
303 struct in6_addr mcaddr;
304 struct in6_addr *target;
306 neigh->updated = jiffies;
307 read_unlock_bh(&neigh->lock);
309 target = (struct in6_addr *)&neigh->primary_key;
310 addrconf_addr_solict_mult(target, &mcaddr);
311 ndisc_send_ns(rt->rt6i_dev, NULL, target, &mcaddr, NULL);
313 read_unlock_bh(&neigh->lock);
316 static inline void rt6_probe(struct rt6_info *rt)
323 * Default Router Selection (RFC 2461 6.3.6)
325 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
327 struct net_device *dev = rt->rt6i_dev;
328 if (!oif || dev->ifindex == oif)
330 if ((dev->flags & IFF_LOOPBACK) &&
331 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
336 static inline int rt6_check_neigh(struct rt6_info *rt)
338 struct neighbour *neigh = rt->rt6i_nexthop;
340 if (rt->rt6i_flags & RTF_NONEXTHOP ||
341 !(rt->rt6i_flags & RTF_GATEWAY))
344 read_lock_bh(&neigh->lock);
345 if (neigh->nud_state & NUD_VALID)
347 #ifdef CONFIG_IPV6_ROUTER_PREF
348 else if (neigh->nud_state & NUD_FAILED)
353 read_unlock_bh(&neigh->lock);
359 static int rt6_score_route(struct rt6_info *rt, int oif,
364 m = rt6_check_dev(rt, oif);
365 if (!m && (strict & RT6_LOOKUP_F_IFACE))
367 #ifdef CONFIG_IPV6_ROUTER_PREF
368 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
370 n = rt6_check_neigh(rt);
371 if (!n && (strict & RT6_LOOKUP_F_REACHABLE))
376 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
377 int *mpri, struct rt6_info *match)
381 if (rt6_check_expired(rt))
384 m = rt6_score_route(rt, oif, strict);
389 if (strict & RT6_LOOKUP_F_REACHABLE)
393 } else if (strict & RT6_LOOKUP_F_REACHABLE) {
401 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
402 struct rt6_info *rr_head,
403 u32 metric, int oif, int strict)
405 struct rt6_info *rt, *match;
409 for (rt = rr_head; rt && rt->rt6i_metric == metric;
410 rt = rt->u.dst.rt6_next)
411 match = find_match(rt, oif, strict, &mpri, match);
412 for (rt = fn->leaf; rt && rt != rr_head && rt->rt6i_metric == metric;
413 rt = rt->u.dst.rt6_next)
414 match = find_match(rt, oif, strict, &mpri, match);
419 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
421 struct rt6_info *match, *rt0;
424 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
425 __func__, fn->leaf, oif);
429 fn->rr_ptr = rt0 = fn->leaf;
431 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict);
434 (strict & RT6_LOOKUP_F_REACHABLE)) {
435 struct rt6_info *next = rt0->u.dst.rt6_next;
437 /* no entries matched; do round-robin */
438 if (!next || next->rt6i_metric != rt0->rt6i_metric)
445 RT6_TRACE("%s() => %p\n",
448 net = dev_net(rt0->rt6i_dev);
449 return (match ? match : net->ipv6.ip6_null_entry);
452 #ifdef CONFIG_IPV6_ROUTE_INFO
453 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
454 struct in6_addr *gwaddr)
456 struct net *net = dev_net(dev);
457 struct route_info *rinfo = (struct route_info *) opt;
458 struct in6_addr prefix_buf, *prefix;
460 unsigned long lifetime;
463 if (len < sizeof(struct route_info)) {
467 /* Sanity check for prefix_len and length */
468 if (rinfo->length > 3) {
470 } else if (rinfo->prefix_len > 128) {
472 } else if (rinfo->prefix_len > 64) {
473 if (rinfo->length < 2) {
476 } else if (rinfo->prefix_len > 0) {
477 if (rinfo->length < 1) {
482 pref = rinfo->route_pref;
483 if (pref == ICMPV6_ROUTER_PREF_INVALID)
486 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
488 if (rinfo->length == 3)
489 prefix = (struct in6_addr *)rinfo->prefix;
491 /* this function is safe */
492 ipv6_addr_prefix(&prefix_buf,
493 (struct in6_addr *)rinfo->prefix,
495 prefix = &prefix_buf;
498 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len, gwaddr,
501 if (rt && !lifetime) {
507 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
510 rt->rt6i_flags = RTF_ROUTEINFO |
511 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
514 if (!addrconf_finite_timeout(lifetime)) {
515 rt->rt6i_flags &= ~RTF_EXPIRES;
517 rt->rt6i_expires = jiffies + HZ * lifetime;
518 rt->rt6i_flags |= RTF_EXPIRES;
520 dst_release(&rt->u.dst);
526 #define BACKTRACK(__net, saddr) \
528 if (rt == __net->ipv6.ip6_null_entry) { \
529 struct fib6_node *pn; \
531 if (fn->fn_flags & RTN_TL_ROOT) \
534 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
535 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
538 if (fn->fn_flags & RTN_RTINFO) \
544 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
545 struct fib6_table *table,
546 struct flowi *fl, int flags)
548 struct fib6_node *fn;
551 read_lock_bh(&table->tb6_lock);
552 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
555 rt = rt6_device_match(net, rt, &fl->fl6_src, fl->oif, flags);
556 BACKTRACK(net, &fl->fl6_src);
558 dst_use(&rt->u.dst, jiffies);
559 read_unlock_bh(&table->tb6_lock);
564 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
565 const struct in6_addr *saddr, int oif, int strict)
575 struct dst_entry *dst;
576 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
579 memcpy(&fl.fl6_src, saddr, sizeof(*saddr));
580 flags |= RT6_LOOKUP_F_HAS_SADDR;
583 dst = fib6_rule_lookup(net, &fl, flags, ip6_pol_route_lookup);
585 return (struct rt6_info *) dst;
592 EXPORT_SYMBOL(rt6_lookup);
594 /* ip6_ins_rt is called with FREE table->tb6_lock.
595 It takes new route entry, the addition fails by any reason the
596 route is freed. In any case, if caller does not hold it, it may
600 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info)
603 struct fib6_table *table;
605 table = rt->rt6i_table;
606 write_lock_bh(&table->tb6_lock);
607 err = fib6_add(&table->tb6_root, rt, info);
608 write_unlock_bh(&table->tb6_lock);
613 int ip6_ins_rt(struct rt6_info *rt)
615 struct nl_info info = {
616 .nl_net = dev_net(rt->rt6i_dev),
618 return __ip6_ins_rt(rt, &info);
621 static struct rt6_info *rt6_alloc_cow(struct rt6_info *ort, struct in6_addr *daddr,
622 struct in6_addr *saddr)
630 rt = ip6_rt_copy(ort);
633 struct neighbour *neigh;
634 int attempts = !in_softirq();
636 if (!(rt->rt6i_flags&RTF_GATEWAY)) {
637 if (rt->rt6i_dst.plen != 128 &&
638 ipv6_addr_equal(&rt->rt6i_dst.addr, daddr))
639 rt->rt6i_flags |= RTF_ANYCAST;
640 ipv6_addr_copy(&rt->rt6i_gateway, daddr);
643 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
644 rt->rt6i_dst.plen = 128;
645 rt->rt6i_flags |= RTF_CACHE;
646 rt->u.dst.flags |= DST_HOST;
648 #ifdef CONFIG_IPV6_SUBTREES
649 if (rt->rt6i_src.plen && saddr) {
650 ipv6_addr_copy(&rt->rt6i_src.addr, saddr);
651 rt->rt6i_src.plen = 128;
656 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
658 struct net *net = dev_net(rt->rt6i_dev);
659 int saved_rt_min_interval =
660 net->ipv6.sysctl.ip6_rt_gc_min_interval;
661 int saved_rt_elasticity =
662 net->ipv6.sysctl.ip6_rt_gc_elasticity;
664 if (attempts-- > 0) {
665 net->ipv6.sysctl.ip6_rt_gc_elasticity = 1;
666 net->ipv6.sysctl.ip6_rt_gc_min_interval = 0;
668 ip6_dst_gc(&net->ipv6.ip6_dst_ops);
670 net->ipv6.sysctl.ip6_rt_gc_elasticity =
672 net->ipv6.sysctl.ip6_rt_gc_min_interval =
673 saved_rt_min_interval;
679 "Neighbour table overflow.\n");
680 dst_free(&rt->u.dst);
683 rt->rt6i_nexthop = neigh;
690 static struct rt6_info *rt6_alloc_clone(struct rt6_info *ort, struct in6_addr *daddr)
692 struct rt6_info *rt = ip6_rt_copy(ort);
694 ipv6_addr_copy(&rt->rt6i_dst.addr, daddr);
695 rt->rt6i_dst.plen = 128;
696 rt->rt6i_flags |= RTF_CACHE;
697 rt->u.dst.flags |= DST_HOST;
698 rt->rt6i_nexthop = neigh_clone(ort->rt6i_nexthop);
703 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
704 struct flowi *fl, int flags)
706 struct fib6_node *fn;
707 struct rt6_info *rt, *nrt;
711 int reachable = net->ipv6.devconf_all->forwarding ? 0 : RT6_LOOKUP_F_REACHABLE;
713 strict |= flags & RT6_LOOKUP_F_IFACE;
716 read_lock_bh(&table->tb6_lock);
719 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
722 rt = rt6_select(fn, oif, strict | reachable);
724 BACKTRACK(net, &fl->fl6_src);
725 if (rt == net->ipv6.ip6_null_entry ||
726 rt->rt6i_flags & RTF_CACHE)
729 dst_hold(&rt->u.dst);
730 read_unlock_bh(&table->tb6_lock);
732 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
733 nrt = rt6_alloc_cow(rt, &fl->fl6_dst, &fl->fl6_src);
735 #if CLONE_OFFLINK_ROUTE
736 nrt = rt6_alloc_clone(rt, &fl->fl6_dst);
742 dst_release(&rt->u.dst);
743 rt = nrt ? : net->ipv6.ip6_null_entry;
745 dst_hold(&rt->u.dst);
747 err = ip6_ins_rt(nrt);
756 * Race condition! In the gap, when table->tb6_lock was
757 * released someone could insert this route. Relookup.
759 dst_release(&rt->u.dst);
767 dst_hold(&rt->u.dst);
768 read_unlock_bh(&table->tb6_lock);
770 rt->u.dst.lastuse = jiffies;
776 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
777 struct flowi *fl, int flags)
779 return ip6_pol_route(net, table, fl->iif, fl, flags);
782 void ip6_route_input(struct sk_buff *skb)
784 struct ipv6hdr *iph = ipv6_hdr(skb);
785 struct net *net = dev_net(skb->dev);
786 int flags = RT6_LOOKUP_F_HAS_SADDR;
788 .iif = skb->dev->ifindex,
793 .flowlabel = (* (__be32 *) iph)&IPV6_FLOWINFO_MASK,
797 .proto = iph->nexthdr,
800 if (rt6_need_strict(&iph->daddr) && skb->dev->type != ARPHRD_PIMREG)
801 flags |= RT6_LOOKUP_F_IFACE;
803 skb_dst_set(skb, fib6_rule_lookup(net, &fl, flags, ip6_pol_route_input));
806 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
807 struct flowi *fl, int flags)
809 return ip6_pol_route(net, table, fl->oif, fl, flags);
812 struct dst_entry * ip6_route_output(struct net *net, struct sock *sk,
817 if (rt6_need_strict(&fl->fl6_dst))
818 flags |= RT6_LOOKUP_F_IFACE;
820 if (!ipv6_addr_any(&fl->fl6_src))
821 flags |= RT6_LOOKUP_F_HAS_SADDR;
823 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
825 return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
828 EXPORT_SYMBOL(ip6_route_output);
830 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
832 struct rt6_info *ort = (struct rt6_info *) *dstp;
833 struct rt6_info *rt = (struct rt6_info *)
834 dst_alloc(&ip6_dst_blackhole_ops);
835 struct dst_entry *new = NULL;
840 atomic_set(&new->__refcnt, 1);
842 new->input = dst_discard;
843 new->output = dst_discard;
845 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
846 new->dev = ort->u.dst.dev;
849 rt->rt6i_idev = ort->rt6i_idev;
851 in6_dev_hold(rt->rt6i_idev);
852 rt->rt6i_expires = 0;
854 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
855 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
858 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
859 #ifdef CONFIG_IPV6_SUBTREES
860 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
868 return (new ? 0 : -ENOMEM);
870 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
873 * Destination cache support functions
876 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
880 rt = (struct rt6_info *) dst;
882 if (rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
888 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
890 struct rt6_info *rt = (struct rt6_info *) dst;
893 if (rt->rt6i_flags & RTF_CACHE) {
894 if (rt6_check_expired(rt)) {
906 static void ip6_link_failure(struct sk_buff *skb)
910 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
912 rt = (struct rt6_info *) skb_dst(skb);
914 if (rt->rt6i_flags&RTF_CACHE) {
915 dst_set_expires(&rt->u.dst, 0);
916 rt->rt6i_flags |= RTF_EXPIRES;
917 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
918 rt->rt6i_node->fn_sernum = -1;
922 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
924 struct rt6_info *rt6 = (struct rt6_info*)dst;
926 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
927 rt6->rt6i_flags |= RTF_MODIFIED;
928 if (mtu < IPV6_MIN_MTU) {
930 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
932 dst->metrics[RTAX_MTU-1] = mtu;
933 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
937 static int ipv6_get_mtu(struct net_device *dev);
939 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
941 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
943 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
944 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
947 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
948 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
949 * IPV6_MAXPLEN is also valid and means: "any MSS,
950 * rely only on pmtu discovery"
952 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
957 static struct dst_entry *icmp6_dst_gc_list;
958 static DEFINE_SPINLOCK(icmp6_dst_lock);
960 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
961 struct neighbour *neigh,
962 const struct in6_addr *addr)
965 struct inet6_dev *idev = in6_dev_get(dev);
966 struct net *net = dev_net(dev);
968 if (unlikely(idev == NULL))
971 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
972 if (unlikely(rt == NULL)) {
981 neigh = ndisc_get_neigh(dev, addr);
987 rt->rt6i_idev = idev;
988 rt->rt6i_nexthop = neigh;
989 atomic_set(&rt->u.dst.__refcnt, 1);
990 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
991 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
992 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
993 rt->u.dst.output = ip6_output;
995 #if 0 /* there's no chance to use these for ndisc */
996 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
999 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1000 rt->rt6i_dst.plen = 128;
1003 spin_lock_bh(&icmp6_dst_lock);
1004 rt->u.dst.next = icmp6_dst_gc_list;
1005 icmp6_dst_gc_list = &rt->u.dst;
1006 spin_unlock_bh(&icmp6_dst_lock);
1008 fib6_force_start_gc(net);
1014 int icmp6_dst_gc(void)
1016 struct dst_entry *dst, *next, **pprev;
1021 spin_lock_bh(&icmp6_dst_lock);
1022 pprev = &icmp6_dst_gc_list;
1024 while ((dst = *pprev) != NULL) {
1025 if (!atomic_read(&dst->__refcnt)) {
1034 spin_unlock_bh(&icmp6_dst_lock);
1039 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1042 struct dst_entry *dst, **pprev;
1044 spin_lock_bh(&icmp6_dst_lock);
1045 pprev = &icmp6_dst_gc_list;
1046 while ((dst = *pprev) != NULL) {
1047 struct rt6_info *rt = (struct rt6_info *) dst;
1048 if (func(rt, arg)) {
1055 spin_unlock_bh(&icmp6_dst_lock);
1058 static int ip6_dst_gc(struct dst_ops *ops)
1060 unsigned long now = jiffies;
1061 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1062 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1063 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1064 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1065 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1066 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1068 if (time_after(rt_last_gc + rt_min_interval, now) &&
1069 atomic_read(&ops->entries) <= rt_max_size)
1072 net->ipv6.ip6_rt_gc_expire++;
1073 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1074 net->ipv6.ip6_rt_last_gc = now;
1075 if (atomic_read(&ops->entries) < ops->gc_thresh)
1076 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1078 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1079 return (atomic_read(&ops->entries) > rt_max_size);
1082 /* Clean host part of a prefix. Not necessary in radix tree,
1083 but results in cleaner routing tables.
1085 Remove it only when all the things will work!
1088 static int ipv6_get_mtu(struct net_device *dev)
1090 int mtu = IPV6_MIN_MTU;
1091 struct inet6_dev *idev;
1093 idev = in6_dev_get(dev);
1095 mtu = idev->cnf.mtu6;
1101 int ip6_dst_hoplimit(struct dst_entry *dst)
1103 int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1105 struct net_device *dev = dst->dev;
1106 struct inet6_dev *idev = in6_dev_get(dev);
1108 hoplimit = idev->cnf.hop_limit;
1111 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1120 int ip6_route_add(struct fib6_config *cfg)
1123 struct net *net = cfg->fc_nlinfo.nl_net;
1124 struct rt6_info *rt = NULL;
1125 struct net_device *dev = NULL;
1126 struct inet6_dev *idev = NULL;
1127 struct fib6_table *table;
1130 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1132 #ifndef CONFIG_IPV6_SUBTREES
1133 if (cfg->fc_src_len)
1136 if (cfg->fc_ifindex) {
1138 dev = dev_get_by_index(net, cfg->fc_ifindex);
1141 idev = in6_dev_get(dev);
1146 if (cfg->fc_metric == 0)
1147 cfg->fc_metric = IP6_RT_PRIO_USER;
1149 table = fib6_new_table(net, cfg->fc_table);
1150 if (table == NULL) {
1155 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1162 rt->u.dst.obsolete = -1;
1163 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1164 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1167 if (cfg->fc_protocol == RTPROT_UNSPEC)
1168 cfg->fc_protocol = RTPROT_BOOT;
1169 rt->rt6i_protocol = cfg->fc_protocol;
1171 addr_type = ipv6_addr_type(&cfg->fc_dst);
1173 if (addr_type & IPV6_ADDR_MULTICAST)
1174 rt->u.dst.input = ip6_mc_input;
1176 rt->u.dst.input = ip6_forward;
1178 rt->u.dst.output = ip6_output;
1180 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1181 rt->rt6i_dst.plen = cfg->fc_dst_len;
1182 if (rt->rt6i_dst.plen == 128)
1183 rt->u.dst.flags = DST_HOST;
1185 #ifdef CONFIG_IPV6_SUBTREES
1186 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1187 rt->rt6i_src.plen = cfg->fc_src_len;
1190 rt->rt6i_metric = cfg->fc_metric;
1192 /* We cannot add true routes via loopback here,
1193 they would result in kernel looping; promote them to reject routes
1195 if ((cfg->fc_flags & RTF_REJECT) ||
1196 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
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->u.dst.output = ip6_pkt_discard_out;
1212 rt->u.dst.input = ip6_pkt_discard;
1213 rt->u.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->u.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->u.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->u.dst.metrics[type - 1] = nla_get_u32(nla);
1302 if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
1303 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1304 if (!dst_mtu(&rt->u.dst))
1305 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1306 if (!dst_metric(&rt->u.dst, RTAX_ADVMSS))
1307 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1308 rt->u.dst.dev = dev;
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);
1322 dst_free(&rt->u.dst);
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->u.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->u.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)
1382 dst_hold(&rt->u.dst);
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->u.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);
1448 dst_hold(&rt->u.dst);
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->u.dst);
1518 /* Duplicate redirect: silently ignore. */
1519 if (neigh == rt->u.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->u.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->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1538 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1539 dst_mtu(&nrt->u.dst));
1541 if (ip6_ins_rt(nrt))
1544 netevent.old = &rt->u.dst;
1545 netevent.new = &nrt->u.dst;
1546 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1548 if (rt->rt6i_flags&RTF_CACHE) {
1554 dst_release(&rt->u.dst);
1559 * Handle ICMP "packet too big" messages
1560 * i.e. Path MTU discovery
1563 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1564 struct net_device *dev, u32 pmtu)
1566 struct rt6_info *rt, *nrt;
1567 struct net *net = dev_net(dev);
1570 rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1574 if (pmtu >= dst_mtu(&rt->u.dst))
1577 if (pmtu < IPV6_MIN_MTU) {
1579 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1580 * MTU (1280) and a fragment header should always be included
1581 * after a node receiving Too Big message reporting PMTU is
1582 * less than the IPv6 Minimum Link MTU.
1584 pmtu = IPV6_MIN_MTU;
1588 /* New mtu received -> path was valid.
1589 They are sent only in response to data packets,
1590 so that this nexthop apparently is reachable. --ANK
1592 dst_confirm(&rt->u.dst);
1594 /* Host route. If it is static, it would be better
1595 not to override it, but add new one, so that
1596 when cache entry will expire old pmtu
1597 would return automatically.
1599 if (rt->rt6i_flags & RTF_CACHE) {
1600 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1602 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1603 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1604 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1609 Two cases are possible:
1610 1. It is connected route. Action: COW
1611 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1613 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1614 nrt = rt6_alloc_cow(rt, daddr, saddr);
1616 nrt = rt6_alloc_clone(rt, daddr);
1619 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1621 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1623 /* According to RFC 1981, detecting PMTU increase shouldn't be
1624 * happened within 5 mins, the recommended timer is 10 mins.
1625 * Here this route expiration time is set to ip6_rt_mtu_expires
1626 * which is 10 mins. After 10 mins the decreased pmtu is expired
1627 * and detecting PMTU increase will be automatically happened.
1629 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1630 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1635 dst_release(&rt->u.dst);
1639 * Misc support functions
1642 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1644 struct net *net = dev_net(ort->rt6i_dev);
1645 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1648 rt->u.dst.input = ort->u.dst.input;
1649 rt->u.dst.output = ort->u.dst.output;
1651 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1652 rt->u.dst.error = ort->u.dst.error;
1653 rt->u.dst.dev = ort->u.dst.dev;
1655 dev_hold(rt->u.dst.dev);
1656 rt->rt6i_idev = ort->rt6i_idev;
1658 in6_dev_hold(rt->rt6i_idev);
1659 rt->u.dst.lastuse = jiffies;
1660 rt->rt6i_expires = 0;
1662 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1663 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1664 rt->rt6i_metric = 0;
1666 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1667 #ifdef CONFIG_IPV6_SUBTREES
1668 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1670 rt->rt6i_table = ort->rt6i_table;
1675 #ifdef CONFIG_IPV6_ROUTE_INFO
1676 static struct rt6_info *rt6_get_route_info(struct net *net,
1677 struct in6_addr *prefix, int prefixlen,
1678 struct in6_addr *gwaddr, int ifindex)
1680 struct fib6_node *fn;
1681 struct rt6_info *rt = NULL;
1682 struct fib6_table *table;
1684 table = fib6_get_table(net, RT6_TABLE_INFO);
1688 write_lock_bh(&table->tb6_lock);
1689 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1693 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1694 if (rt->rt6i_dev->ifindex != ifindex)
1696 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1698 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1700 dst_hold(&rt->u.dst);
1704 write_unlock_bh(&table->tb6_lock);
1708 static struct rt6_info *rt6_add_route_info(struct net *net,
1709 struct in6_addr *prefix, int prefixlen,
1710 struct in6_addr *gwaddr, int ifindex,
1713 struct fib6_config cfg = {
1714 .fc_table = RT6_TABLE_INFO,
1715 .fc_metric = IP6_RT_PRIO_USER,
1716 .fc_ifindex = ifindex,
1717 .fc_dst_len = prefixlen,
1718 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1719 RTF_UP | RTF_PREF(pref),
1721 .fc_nlinfo.nlh = NULL,
1722 .fc_nlinfo.nl_net = net,
1725 ipv6_addr_copy(&cfg.fc_dst, prefix);
1726 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1728 /* We should treat it as a default route if prefix length is 0. */
1730 cfg.fc_flags |= RTF_DEFAULT;
1732 ip6_route_add(&cfg);
1734 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1738 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1740 struct rt6_info *rt;
1741 struct fib6_table *table;
1743 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1747 write_lock_bh(&table->tb6_lock);
1748 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1749 if (dev == rt->rt6i_dev &&
1750 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1751 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1755 dst_hold(&rt->u.dst);
1756 write_unlock_bh(&table->tb6_lock);
1760 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1761 struct net_device *dev,
1764 struct fib6_config cfg = {
1765 .fc_table = RT6_TABLE_DFLT,
1766 .fc_metric = IP6_RT_PRIO_USER,
1767 .fc_ifindex = dev->ifindex,
1768 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1769 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1771 .fc_nlinfo.nlh = NULL,
1772 .fc_nlinfo.nl_net = dev_net(dev),
1775 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1777 ip6_route_add(&cfg);
1779 return rt6_get_dflt_router(gwaddr, dev);
1782 void rt6_purge_dflt_routers(struct net *net)
1784 struct rt6_info *rt;
1785 struct fib6_table *table;
1787 /* NOTE: Keep consistent with rt6_get_dflt_router */
1788 table = fib6_get_table(net, RT6_TABLE_DFLT);
1793 read_lock_bh(&table->tb6_lock);
1794 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1795 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1796 dst_hold(&rt->u.dst);
1797 read_unlock_bh(&table->tb6_lock);
1802 read_unlock_bh(&table->tb6_lock);
1805 static void rtmsg_to_fib6_config(struct net *net,
1806 struct in6_rtmsg *rtmsg,
1807 struct fib6_config *cfg)
1809 memset(cfg, 0, sizeof(*cfg));
1811 cfg->fc_table = RT6_TABLE_MAIN;
1812 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1813 cfg->fc_metric = rtmsg->rtmsg_metric;
1814 cfg->fc_expires = rtmsg->rtmsg_info;
1815 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1816 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1817 cfg->fc_flags = rtmsg->rtmsg_flags;
1819 cfg->fc_nlinfo.nl_net = net;
1821 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1822 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1823 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1826 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1828 struct fib6_config cfg;
1829 struct in6_rtmsg rtmsg;
1833 case SIOCADDRT: /* Add a route */
1834 case SIOCDELRT: /* Delete a route */
1835 if (!capable(CAP_NET_ADMIN))
1837 err = copy_from_user(&rtmsg, arg,
1838 sizeof(struct in6_rtmsg));
1842 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1847 err = ip6_route_add(&cfg);
1850 err = ip6_route_del(&cfg);
1864 * Drop the packet on the floor
1867 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1870 struct dst_entry *dst = skb_dst(skb);
1871 switch (ipstats_mib_noroutes) {
1872 case IPSTATS_MIB_INNOROUTES:
1873 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1874 if (type == IPV6_ADDR_ANY) {
1875 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1876 IPSTATS_MIB_INADDRERRORS);
1880 case IPSTATS_MIB_OUTNOROUTES:
1881 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1882 ipstats_mib_noroutes);
1885 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1890 static int ip6_pkt_discard(struct sk_buff *skb)
1892 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1895 static int ip6_pkt_discard_out(struct sk_buff *skb)
1897 skb->dev = skb_dst(skb)->dev;
1898 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1901 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1903 static int ip6_pkt_prohibit(struct sk_buff *skb)
1905 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1908 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1910 skb->dev = skb_dst(skb)->dev;
1911 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1917 * Allocate a dst for local (unicast / anycast) address.
1920 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1921 const struct in6_addr *addr,
1924 struct net *net = dev_net(idev->dev);
1925 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1926 struct neighbour *neigh;
1929 return ERR_PTR(-ENOMEM);
1931 dev_hold(net->loopback_dev);
1934 rt->u.dst.flags = DST_HOST;
1935 rt->u.dst.input = ip6_input;
1936 rt->u.dst.output = ip6_output;
1937 rt->rt6i_dev = net->loopback_dev;
1938 rt->rt6i_idev = idev;
1939 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1940 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1941 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1942 rt->u.dst.obsolete = -1;
1944 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1946 rt->rt6i_flags |= RTF_ANYCAST;
1948 rt->rt6i_flags |= RTF_LOCAL;
1949 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1950 if (IS_ERR(neigh)) {
1951 dst_free(&rt->u.dst);
1953 /* We are casting this because that is the return
1954 * value type. But an errno encoded pointer is the
1955 * same regardless of the underlying pointer type,
1956 * and that's what we are returning. So this is OK.
1958 return (struct rt6_info *) neigh;
1960 rt->rt6i_nexthop = neigh;
1962 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1963 rt->rt6i_dst.plen = 128;
1964 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1966 atomic_set(&rt->u.dst.__refcnt, 1);
1971 struct arg_dev_net {
1972 struct net_device *dev;
1976 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1978 struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1979 struct net *net = ((struct arg_dev_net *)arg)->net;
1981 if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1982 rt != net->ipv6.ip6_null_entry) {
1983 RT6_TRACE("deleted by ifdown %p\n", rt);
1989 void rt6_ifdown(struct net *net, struct net_device *dev)
1991 struct arg_dev_net adn = {
1996 fib6_clean_all(net, fib6_ifdown, 0, &adn);
1997 icmp6_clean_all(fib6_ifdown, &adn);
2000 struct rt6_mtu_change_arg
2002 struct net_device *dev;
2006 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2008 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2009 struct inet6_dev *idev;
2010 struct net *net = dev_net(arg->dev);
2012 /* In IPv6 pmtu discovery is not optional,
2013 so that RTAX_MTU lock cannot disable it.
2014 We still use this lock to block changes
2015 caused by addrconf/ndisc.
2018 idev = __in6_dev_get(arg->dev);
2022 /* For administrative MTU increase, there is no way to discover
2023 IPv6 PMTU increase, so PMTU increase should be updated here.
2024 Since RFC 1981 doesn't include administrative MTU increase
2025 update PMTU increase is a MUST. (i.e. jumbo frame)
2028 If new MTU is less than route PMTU, this new MTU will be the
2029 lowest MTU in the path, update the route PMTU to reflect PMTU
2030 decreases; if new MTU is greater than route PMTU, and the
2031 old MTU is the lowest MTU in the path, update the route PMTU
2032 to reflect the increase. In this case if the other nodes' MTU
2033 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2036 if (rt->rt6i_dev == arg->dev &&
2037 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
2038 (dst_mtu(&rt->u.dst) >= arg->mtu ||
2039 (dst_mtu(&rt->u.dst) < arg->mtu &&
2040 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
2041 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
2042 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
2047 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2049 struct rt6_mtu_change_arg arg = {
2054 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2057 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2058 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2059 [RTA_OIF] = { .type = NLA_U32 },
2060 [RTA_IIF] = { .type = NLA_U32 },
2061 [RTA_PRIORITY] = { .type = NLA_U32 },
2062 [RTA_METRICS] = { .type = NLA_NESTED },
2065 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2066 struct fib6_config *cfg)
2069 struct nlattr *tb[RTA_MAX+1];
2072 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2077 rtm = nlmsg_data(nlh);
2078 memset(cfg, 0, sizeof(*cfg));
2080 cfg->fc_table = rtm->rtm_table;
2081 cfg->fc_dst_len = rtm->rtm_dst_len;
2082 cfg->fc_src_len = rtm->rtm_src_len;
2083 cfg->fc_flags = RTF_UP;
2084 cfg->fc_protocol = rtm->rtm_protocol;
2086 if (rtm->rtm_type == RTN_UNREACHABLE)
2087 cfg->fc_flags |= RTF_REJECT;
2089 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2090 cfg->fc_nlinfo.nlh = nlh;
2091 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2093 if (tb[RTA_GATEWAY]) {
2094 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2095 cfg->fc_flags |= RTF_GATEWAY;
2099 int plen = (rtm->rtm_dst_len + 7) >> 3;
2101 if (nla_len(tb[RTA_DST]) < plen)
2104 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2108 int plen = (rtm->rtm_src_len + 7) >> 3;
2110 if (nla_len(tb[RTA_SRC]) < plen)
2113 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2117 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2119 if (tb[RTA_PRIORITY])
2120 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2122 if (tb[RTA_METRICS]) {
2123 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2124 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2128 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2135 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2137 struct fib6_config cfg;
2140 err = rtm_to_fib6_config(skb, nlh, &cfg);
2144 return ip6_route_del(&cfg);
2147 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2149 struct fib6_config cfg;
2152 err = rtm_to_fib6_config(skb, nlh, &cfg);
2156 return ip6_route_add(&cfg);
2159 static inline size_t rt6_nlmsg_size(void)
2161 return NLMSG_ALIGN(sizeof(struct rtmsg))
2162 + nla_total_size(16) /* RTA_SRC */
2163 + nla_total_size(16) /* RTA_DST */
2164 + nla_total_size(16) /* RTA_GATEWAY */
2165 + nla_total_size(16) /* RTA_PREFSRC */
2166 + nla_total_size(4) /* RTA_TABLE */
2167 + nla_total_size(4) /* RTA_IIF */
2168 + nla_total_size(4) /* RTA_OIF */
2169 + nla_total_size(4) /* RTA_PRIORITY */
2170 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2171 + nla_total_size(sizeof(struct rta_cacheinfo));
2174 static int rt6_fill_node(struct net *net,
2175 struct sk_buff *skb, struct rt6_info *rt,
2176 struct in6_addr *dst, struct in6_addr *src,
2177 int iif, int type, u32 pid, u32 seq,
2178 int prefix, int nowait, unsigned int flags)
2181 struct nlmsghdr *nlh;
2185 if (prefix) { /* user wants prefix routes only */
2186 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2187 /* success since this is not a prefix route */
2192 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2196 rtm = nlmsg_data(nlh);
2197 rtm->rtm_family = AF_INET6;
2198 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2199 rtm->rtm_src_len = rt->rt6i_src.plen;
2202 table = rt->rt6i_table->tb6_id;
2204 table = RT6_TABLE_UNSPEC;
2205 rtm->rtm_table = table;
2206 NLA_PUT_U32(skb, RTA_TABLE, table);
2207 if (rt->rt6i_flags&RTF_REJECT)
2208 rtm->rtm_type = RTN_UNREACHABLE;
2209 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2210 rtm->rtm_type = RTN_LOCAL;
2212 rtm->rtm_type = RTN_UNICAST;
2214 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2215 rtm->rtm_protocol = rt->rt6i_protocol;
2216 if (rt->rt6i_flags&RTF_DYNAMIC)
2217 rtm->rtm_protocol = RTPROT_REDIRECT;
2218 else if (rt->rt6i_flags & RTF_ADDRCONF)
2219 rtm->rtm_protocol = RTPROT_KERNEL;
2220 else if (rt->rt6i_flags&RTF_DEFAULT)
2221 rtm->rtm_protocol = RTPROT_RA;
2223 if (rt->rt6i_flags&RTF_CACHE)
2224 rtm->rtm_flags |= RTM_F_CLONED;
2227 NLA_PUT(skb, RTA_DST, 16, dst);
2228 rtm->rtm_dst_len = 128;
2229 } else if (rtm->rtm_dst_len)
2230 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2231 #ifdef CONFIG_IPV6_SUBTREES
2233 NLA_PUT(skb, RTA_SRC, 16, src);
2234 rtm->rtm_src_len = 128;
2235 } else if (rtm->rtm_src_len)
2236 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2239 #ifdef CONFIG_IPV6_MROUTE
2240 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2241 int err = ip6mr_get_route(net, skb, rtm, nowait);
2246 goto nla_put_failure;
2248 if (err == -EMSGSIZE)
2249 goto nla_put_failure;
2254 NLA_PUT_U32(skb, RTA_IIF, iif);
2256 struct inet6_dev *idev = ip6_dst_idev(&rt->u.dst);
2257 struct in6_addr saddr_buf;
2258 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2259 dst, 0, &saddr_buf) == 0)
2260 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2263 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2264 goto nla_put_failure;
2266 if (rt->u.dst.neighbour)
2267 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2270 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2272 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2274 if (!(rt->rt6i_flags & RTF_EXPIRES))
2276 else if (rt->rt6i_expires - jiffies < INT_MAX)
2277 expires = rt->rt6i_expires - jiffies;
2281 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2282 expires, rt->u.dst.error) < 0)
2283 goto nla_put_failure;
2285 return nlmsg_end(skb, nlh);
2288 nlmsg_cancel(skb, nlh);
2292 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2294 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2297 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2298 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2299 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2303 return rt6_fill_node(arg->net,
2304 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2305 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2306 prefix, 0, NLM_F_MULTI);
2309 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2311 struct net *net = sock_net(in_skb->sk);
2312 struct nlattr *tb[RTA_MAX+1];
2313 struct rt6_info *rt;
2314 struct sk_buff *skb;
2319 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2324 memset(&fl, 0, sizeof(fl));
2327 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2330 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2334 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2337 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2341 iif = nla_get_u32(tb[RTA_IIF]);
2344 fl.oif = nla_get_u32(tb[RTA_OIF]);
2347 struct net_device *dev;
2348 dev = __dev_get_by_index(net, iif);
2355 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2361 /* Reserve room for dummy headers, this skb can pass
2362 through good chunk of routing engine.
2364 skb_reset_mac_header(skb);
2365 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2367 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2368 skb_dst_set(skb, &rt->u.dst);
2370 err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2371 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2372 nlh->nlmsg_seq, 0, 0, 0);
2378 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2383 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2385 struct sk_buff *skb;
2386 struct net *net = info->nl_net;
2391 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2393 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2397 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2398 event, info->pid, seq, 0, 0, 0);
2400 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2401 WARN_ON(err == -EMSGSIZE);
2405 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2406 info->nlh, gfp_any());
2410 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2413 static int ip6_route_dev_notify(struct notifier_block *this,
2414 unsigned long event, void *data)
2416 struct net_device *dev = (struct net_device *)data;
2417 struct net *net = dev_net(dev);
2419 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2420 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2421 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2422 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2423 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2424 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2425 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2426 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2437 #ifdef CONFIG_PROC_FS
2439 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2450 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2452 struct seq_file *m = p_arg;
2454 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2456 #ifdef CONFIG_IPV6_SUBTREES
2457 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2459 seq_puts(m, "00000000000000000000000000000000 00 ");
2462 if (rt->rt6i_nexthop) {
2463 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2465 seq_puts(m, "00000000000000000000000000000000");
2467 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2468 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2469 rt->u.dst.__use, rt->rt6i_flags,
2470 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2474 static int ipv6_route_show(struct seq_file *m, void *v)
2476 struct net *net = (struct net *)m->private;
2477 fib6_clean_all(net, rt6_info_route, 0, m);
2481 static int ipv6_route_open(struct inode *inode, struct file *file)
2483 return single_open_net(inode, file, ipv6_route_show);
2486 static const struct file_operations ipv6_route_proc_fops = {
2487 .owner = THIS_MODULE,
2488 .open = ipv6_route_open,
2490 .llseek = seq_lseek,
2491 .release = single_release_net,
2494 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2496 struct net *net = (struct net *)seq->private;
2497 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2498 net->ipv6.rt6_stats->fib_nodes,
2499 net->ipv6.rt6_stats->fib_route_nodes,
2500 net->ipv6.rt6_stats->fib_rt_alloc,
2501 net->ipv6.rt6_stats->fib_rt_entries,
2502 net->ipv6.rt6_stats->fib_rt_cache,
2503 atomic_read(&net->ipv6.ip6_dst_ops.entries),
2504 net->ipv6.rt6_stats->fib_discarded_routes);
2509 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2511 return single_open_net(inode, file, rt6_stats_seq_show);
2514 static const struct file_operations rt6_stats_seq_fops = {
2515 .owner = THIS_MODULE,
2516 .open = rt6_stats_seq_open,
2518 .llseek = seq_lseek,
2519 .release = single_release_net,
2521 #endif /* CONFIG_PROC_FS */
2523 #ifdef CONFIG_SYSCTL
2526 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2527 void __user *buffer, size_t *lenp, loff_t *ppos)
2529 struct net *net = current->nsproxy->net_ns;
2530 int delay = net->ipv6.sysctl.flush_delay;
2532 proc_dointvec(ctl, write, buffer, lenp, ppos);
2533 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2539 ctl_table ipv6_route_table_template[] = {
2541 .procname = "flush",
2542 .data = &init_net.ipv6.sysctl.flush_delay,
2543 .maxlen = sizeof(int),
2545 .proc_handler = ipv6_sysctl_rtcache_flush
2548 .procname = "gc_thresh",
2549 .data = &ip6_dst_ops_template.gc_thresh,
2550 .maxlen = sizeof(int),
2552 .proc_handler = proc_dointvec,
2555 .procname = "max_size",
2556 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2557 .maxlen = sizeof(int),
2559 .proc_handler = proc_dointvec,
2562 .procname = "gc_min_interval",
2563 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2564 .maxlen = sizeof(int),
2566 .proc_handler = proc_dointvec_jiffies,
2569 .procname = "gc_timeout",
2570 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2571 .maxlen = sizeof(int),
2573 .proc_handler = proc_dointvec_jiffies,
2576 .procname = "gc_interval",
2577 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2578 .maxlen = sizeof(int),
2580 .proc_handler = proc_dointvec_jiffies,
2583 .procname = "gc_elasticity",
2584 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2585 .maxlen = sizeof(int),
2587 .proc_handler = proc_dointvec_jiffies,
2590 .procname = "mtu_expires",
2591 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2592 .maxlen = sizeof(int),
2594 .proc_handler = proc_dointvec_jiffies,
2597 .procname = "min_adv_mss",
2598 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2599 .maxlen = sizeof(int),
2601 .proc_handler = proc_dointvec_jiffies,
2604 .procname = "gc_min_interval_ms",
2605 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2606 .maxlen = sizeof(int),
2608 .proc_handler = proc_dointvec_ms_jiffies,
2613 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2615 struct ctl_table *table;
2617 table = kmemdup(ipv6_route_table_template,
2618 sizeof(ipv6_route_table_template),
2622 table[0].data = &net->ipv6.sysctl.flush_delay;
2623 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2624 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2625 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2626 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2627 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2628 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2629 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2630 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2631 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2638 static int __net_init ip6_route_net_init(struct net *net)
2642 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2643 sizeof(net->ipv6.ip6_dst_ops));
2645 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2646 sizeof(*net->ipv6.ip6_null_entry),
2648 if (!net->ipv6.ip6_null_entry)
2649 goto out_ip6_dst_ops;
2650 net->ipv6.ip6_null_entry->u.dst.path =
2651 (struct dst_entry *)net->ipv6.ip6_null_entry;
2652 net->ipv6.ip6_null_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2654 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2655 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2656 sizeof(*net->ipv6.ip6_prohibit_entry),
2658 if (!net->ipv6.ip6_prohibit_entry)
2659 goto out_ip6_null_entry;
2660 net->ipv6.ip6_prohibit_entry->u.dst.path =
2661 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2662 net->ipv6.ip6_prohibit_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2664 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2665 sizeof(*net->ipv6.ip6_blk_hole_entry),
2667 if (!net->ipv6.ip6_blk_hole_entry)
2668 goto out_ip6_prohibit_entry;
2669 net->ipv6.ip6_blk_hole_entry->u.dst.path =
2670 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2671 net->ipv6.ip6_blk_hole_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2674 net->ipv6.sysctl.flush_delay = 0;
2675 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2676 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2677 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2678 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2679 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2680 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2681 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2683 #ifdef CONFIG_PROC_FS
2684 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2685 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2687 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2693 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2694 out_ip6_prohibit_entry:
2695 kfree(net->ipv6.ip6_prohibit_entry);
2697 kfree(net->ipv6.ip6_null_entry);
2703 static void __net_exit ip6_route_net_exit(struct net *net)
2705 #ifdef CONFIG_PROC_FS
2706 proc_net_remove(net, "ipv6_route");
2707 proc_net_remove(net, "rt6_stats");
2709 kfree(net->ipv6.ip6_null_entry);
2710 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2711 kfree(net->ipv6.ip6_prohibit_entry);
2712 kfree(net->ipv6.ip6_blk_hole_entry);
2716 static struct pernet_operations ip6_route_net_ops = {
2717 .init = ip6_route_net_init,
2718 .exit = ip6_route_net_exit,
2721 static struct notifier_block ip6_route_dev_notifier = {
2722 .notifier_call = ip6_route_dev_notify,
2726 int __init ip6_route_init(void)
2731 ip6_dst_ops_template.kmem_cachep =
2732 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2733 SLAB_HWCACHE_ALIGN, NULL);
2734 if (!ip6_dst_ops_template.kmem_cachep)
2737 ret = register_pernet_subsys(&ip6_route_net_ops);
2739 goto out_kmem_cache;
2741 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2743 /* Registering of the loopback is done before this portion of code,
2744 * the loopback reference in rt6_info will not be taken, do it
2745 * manually for init_net */
2746 init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2747 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2748 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2749 init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2750 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2751 init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2752 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2756 goto out_register_subsys;
2762 ret = fib6_rules_init();
2767 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2768 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2769 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2770 goto fib6_rules_init;
2772 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2774 goto fib6_rules_init;
2780 fib6_rules_cleanup();
2785 out_register_subsys:
2786 unregister_pernet_subsys(&ip6_route_net_ops);
2788 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2792 void ip6_route_cleanup(void)
2794 unregister_netdevice_notifier(&ip6_route_dev_notifier);
2795 fib6_rules_cleanup();
2798 unregister_pernet_subsys(&ip6_route_net_ops);
2799 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);