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 && 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)
901 static void ip6_link_failure(struct sk_buff *skb)
905 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
907 rt = (struct rt6_info *) skb_dst(skb);
909 if (rt->rt6i_flags&RTF_CACHE) {
910 dst_set_expires(&rt->u.dst, 0);
911 rt->rt6i_flags |= RTF_EXPIRES;
912 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
913 rt->rt6i_node->fn_sernum = -1;
917 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
919 struct rt6_info *rt6 = (struct rt6_info*)dst;
921 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
922 rt6->rt6i_flags |= RTF_MODIFIED;
923 if (mtu < IPV6_MIN_MTU) {
925 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
927 dst->metrics[RTAX_MTU-1] = mtu;
928 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
932 static int ipv6_get_mtu(struct net_device *dev);
934 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
936 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
938 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
939 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
942 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
943 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
944 * IPV6_MAXPLEN is also valid and means: "any MSS,
945 * rely only on pmtu discovery"
947 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
952 static struct dst_entry *icmp6_dst_gc_list;
953 static DEFINE_SPINLOCK(icmp6_dst_lock);
955 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
956 struct neighbour *neigh,
957 const struct in6_addr *addr)
960 struct inet6_dev *idev = in6_dev_get(dev);
961 struct net *net = dev_net(dev);
963 if (unlikely(idev == NULL))
966 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
967 if (unlikely(rt == NULL)) {
976 neigh = ndisc_get_neigh(dev, addr);
982 rt->rt6i_idev = idev;
983 rt->rt6i_nexthop = neigh;
984 atomic_set(&rt->u.dst.__refcnt, 1);
985 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
986 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
987 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
988 rt->u.dst.output = ip6_output;
990 #if 0 /* there's no chance to use these for ndisc */
991 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
994 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
995 rt->rt6i_dst.plen = 128;
998 spin_lock_bh(&icmp6_dst_lock);
999 rt->u.dst.next = icmp6_dst_gc_list;
1000 icmp6_dst_gc_list = &rt->u.dst;
1001 spin_unlock_bh(&icmp6_dst_lock);
1003 fib6_force_start_gc(net);
1009 int icmp6_dst_gc(void)
1011 struct dst_entry *dst, *next, **pprev;
1016 spin_lock_bh(&icmp6_dst_lock);
1017 pprev = &icmp6_dst_gc_list;
1019 while ((dst = *pprev) != NULL) {
1020 if (!atomic_read(&dst->__refcnt)) {
1029 spin_unlock_bh(&icmp6_dst_lock);
1034 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1037 struct dst_entry *dst, **pprev;
1039 spin_lock_bh(&icmp6_dst_lock);
1040 pprev = &icmp6_dst_gc_list;
1041 while ((dst = *pprev) != NULL) {
1042 struct rt6_info *rt = (struct rt6_info *) dst;
1043 if (func(rt, arg)) {
1050 spin_unlock_bh(&icmp6_dst_lock);
1053 static int ip6_dst_gc(struct dst_ops *ops)
1055 unsigned long now = jiffies;
1056 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1057 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1058 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1059 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1060 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1061 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1063 if (time_after(rt_last_gc + rt_min_interval, now) &&
1064 atomic_read(&ops->entries) <= rt_max_size)
1067 net->ipv6.ip6_rt_gc_expire++;
1068 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1069 net->ipv6.ip6_rt_last_gc = now;
1070 if (atomic_read(&ops->entries) < ops->gc_thresh)
1071 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1073 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1074 return (atomic_read(&ops->entries) > rt_max_size);
1077 /* Clean host part of a prefix. Not necessary in radix tree,
1078 but results in cleaner routing tables.
1080 Remove it only when all the things will work!
1083 static int ipv6_get_mtu(struct net_device *dev)
1085 int mtu = IPV6_MIN_MTU;
1086 struct inet6_dev *idev;
1088 idev = in6_dev_get(dev);
1090 mtu = idev->cnf.mtu6;
1096 int ip6_dst_hoplimit(struct dst_entry *dst)
1098 int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1100 struct net_device *dev = dst->dev;
1101 struct inet6_dev *idev = in6_dev_get(dev);
1103 hoplimit = idev->cnf.hop_limit;
1106 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1115 int ip6_route_add(struct fib6_config *cfg)
1118 struct net *net = cfg->fc_nlinfo.nl_net;
1119 struct rt6_info *rt = NULL;
1120 struct net_device *dev = NULL;
1121 struct inet6_dev *idev = NULL;
1122 struct fib6_table *table;
1125 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1127 #ifndef CONFIG_IPV6_SUBTREES
1128 if (cfg->fc_src_len)
1131 if (cfg->fc_ifindex) {
1133 dev = dev_get_by_index(net, cfg->fc_ifindex);
1136 idev = in6_dev_get(dev);
1141 if (cfg->fc_metric == 0)
1142 cfg->fc_metric = IP6_RT_PRIO_USER;
1144 table = fib6_new_table(net, cfg->fc_table);
1145 if (table == NULL) {
1150 rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1157 rt->u.dst.obsolete = -1;
1158 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1159 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1162 if (cfg->fc_protocol == RTPROT_UNSPEC)
1163 cfg->fc_protocol = RTPROT_BOOT;
1164 rt->rt6i_protocol = cfg->fc_protocol;
1166 addr_type = ipv6_addr_type(&cfg->fc_dst);
1168 if (addr_type & IPV6_ADDR_MULTICAST)
1169 rt->u.dst.input = ip6_mc_input;
1171 rt->u.dst.input = ip6_forward;
1173 rt->u.dst.output = ip6_output;
1175 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1176 rt->rt6i_dst.plen = cfg->fc_dst_len;
1177 if (rt->rt6i_dst.plen == 128)
1178 rt->u.dst.flags = DST_HOST;
1180 #ifdef CONFIG_IPV6_SUBTREES
1181 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1182 rt->rt6i_src.plen = cfg->fc_src_len;
1185 rt->rt6i_metric = cfg->fc_metric;
1187 /* We cannot add true routes via loopback here,
1188 they would result in kernel looping; promote them to reject routes
1190 if ((cfg->fc_flags & RTF_REJECT) ||
1191 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1192 /* hold loopback dev/idev if we haven't done so. */
1193 if (dev != net->loopback_dev) {
1198 dev = net->loopback_dev;
1200 idev = in6_dev_get(dev);
1206 rt->u.dst.output = ip6_pkt_discard_out;
1207 rt->u.dst.input = ip6_pkt_discard;
1208 rt->u.dst.error = -ENETUNREACH;
1209 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1213 if (cfg->fc_flags & RTF_GATEWAY) {
1214 struct in6_addr *gw_addr;
1217 gw_addr = &cfg->fc_gateway;
1218 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1219 gwa_type = ipv6_addr_type(gw_addr);
1221 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1222 struct rt6_info *grt;
1224 /* IPv6 strictly inhibits using not link-local
1225 addresses as nexthop address.
1226 Otherwise, router will not able to send redirects.
1227 It is very good, but in some (rare!) circumstances
1228 (SIT, PtP, NBMA NOARP links) it is handy to allow
1229 some exceptions. --ANK
1232 if (!(gwa_type&IPV6_ADDR_UNICAST))
1235 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1237 err = -EHOSTUNREACH;
1241 if (dev != grt->rt6i_dev) {
1242 dst_release(&grt->u.dst);
1246 dev = grt->rt6i_dev;
1247 idev = grt->rt6i_idev;
1249 in6_dev_hold(grt->rt6i_idev);
1251 if (!(grt->rt6i_flags&RTF_GATEWAY))
1253 dst_release(&grt->u.dst);
1259 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1267 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1268 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1269 if (IS_ERR(rt->rt6i_nexthop)) {
1270 err = PTR_ERR(rt->rt6i_nexthop);
1271 rt->rt6i_nexthop = NULL;
1276 rt->rt6i_flags = cfg->fc_flags;
1283 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1284 int type = nla_type(nla);
1287 if (type > RTAX_MAX) {
1292 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1297 if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
1298 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1299 if (!dst_mtu(&rt->u.dst))
1300 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1301 if (!dst_metric(&rt->u.dst, RTAX_ADVMSS))
1302 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1303 rt->u.dst.dev = dev;
1304 rt->rt6i_idev = idev;
1305 rt->rt6i_table = table;
1307 cfg->fc_nlinfo.nl_net = dev_net(dev);
1309 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1317 dst_free(&rt->u.dst);
1321 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1324 struct fib6_table *table;
1325 struct net *net = dev_net(rt->rt6i_dev);
1327 if (rt == net->ipv6.ip6_null_entry)
1330 table = rt->rt6i_table;
1331 write_lock_bh(&table->tb6_lock);
1333 err = fib6_del(rt, info);
1334 dst_release(&rt->u.dst);
1336 write_unlock_bh(&table->tb6_lock);
1341 int ip6_del_rt(struct rt6_info *rt)
1343 struct nl_info info = {
1344 .nl_net = dev_net(rt->rt6i_dev),
1346 return __ip6_del_rt(rt, &info);
1349 static int ip6_route_del(struct fib6_config *cfg)
1351 struct fib6_table *table;
1352 struct fib6_node *fn;
1353 struct rt6_info *rt;
1356 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1360 read_lock_bh(&table->tb6_lock);
1362 fn = fib6_locate(&table->tb6_root,
1363 &cfg->fc_dst, cfg->fc_dst_len,
1364 &cfg->fc_src, cfg->fc_src_len);
1367 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1368 if (cfg->fc_ifindex &&
1369 (rt->rt6i_dev == NULL ||
1370 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1372 if (cfg->fc_flags & RTF_GATEWAY &&
1373 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1375 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1377 dst_hold(&rt->u.dst);
1378 read_unlock_bh(&table->tb6_lock);
1380 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1383 read_unlock_bh(&table->tb6_lock);
1391 struct ip6rd_flowi {
1393 struct in6_addr gateway;
1396 static struct rt6_info *__ip6_route_redirect(struct net *net,
1397 struct fib6_table *table,
1401 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1402 struct rt6_info *rt;
1403 struct fib6_node *fn;
1406 * Get the "current" route for this destination and
1407 * check if the redirect has come from approriate router.
1409 * RFC 2461 specifies that redirects should only be
1410 * accepted if they come from the nexthop to the target.
1411 * Due to the way the routes are chosen, this notion
1412 * is a bit fuzzy and one might need to check all possible
1416 read_lock_bh(&table->tb6_lock);
1417 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1419 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1421 * Current route is on-link; redirect is always invalid.
1423 * Seems, previous statement is not true. It could
1424 * be node, which looks for us as on-link (f.e. proxy ndisc)
1425 * But then router serving it might decide, that we should
1426 * know truth 8)8) --ANK (980726).
1428 if (rt6_check_expired(rt))
1430 if (!(rt->rt6i_flags & RTF_GATEWAY))
1432 if (fl->oif != rt->rt6i_dev->ifindex)
1434 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1440 rt = net->ipv6.ip6_null_entry;
1441 BACKTRACK(net, &fl->fl6_src);
1443 dst_hold(&rt->u.dst);
1445 read_unlock_bh(&table->tb6_lock);
1450 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1451 struct in6_addr *src,
1452 struct in6_addr *gateway,
1453 struct net_device *dev)
1455 int flags = RT6_LOOKUP_F_HAS_SADDR;
1456 struct net *net = dev_net(dev);
1457 struct ip6rd_flowi rdfl = {
1459 .oif = dev->ifindex,
1469 ipv6_addr_copy(&rdfl.gateway, gateway);
1471 if (rt6_need_strict(dest))
1472 flags |= RT6_LOOKUP_F_IFACE;
1474 return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1475 flags, __ip6_route_redirect);
1478 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1479 struct in6_addr *saddr,
1480 struct neighbour *neigh, u8 *lladdr, int on_link)
1482 struct rt6_info *rt, *nrt = NULL;
1483 struct netevent_redirect netevent;
1484 struct net *net = dev_net(neigh->dev);
1486 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1488 if (rt == net->ipv6.ip6_null_entry) {
1489 if (net_ratelimit())
1490 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1491 "for redirect target\n");
1496 * We have finally decided to accept it.
1499 neigh_update(neigh, lladdr, NUD_STALE,
1500 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1501 NEIGH_UPDATE_F_OVERRIDE|
1502 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1503 NEIGH_UPDATE_F_ISROUTER))
1507 * Redirect received -> path was valid.
1508 * Look, redirects are sent only in response to data packets,
1509 * so that this nexthop apparently is reachable. --ANK
1511 dst_confirm(&rt->u.dst);
1513 /* Duplicate redirect: silently ignore. */
1514 if (neigh == rt->u.dst.neighbour)
1517 nrt = ip6_rt_copy(rt);
1521 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1523 nrt->rt6i_flags &= ~RTF_GATEWAY;
1525 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1526 nrt->rt6i_dst.plen = 128;
1527 nrt->u.dst.flags |= DST_HOST;
1529 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1530 nrt->rt6i_nexthop = neigh_clone(neigh);
1531 /* Reset pmtu, it may be better */
1532 nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1533 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1534 dst_mtu(&nrt->u.dst));
1536 if (ip6_ins_rt(nrt))
1539 netevent.old = &rt->u.dst;
1540 netevent.new = &nrt->u.dst;
1541 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1543 if (rt->rt6i_flags&RTF_CACHE) {
1549 dst_release(&rt->u.dst);
1554 * Handle ICMP "packet too big" messages
1555 * i.e. Path MTU discovery
1558 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1559 struct net_device *dev, u32 pmtu)
1561 struct rt6_info *rt, *nrt;
1562 struct net *net = dev_net(dev);
1565 rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1569 if (pmtu >= dst_mtu(&rt->u.dst))
1572 if (pmtu < IPV6_MIN_MTU) {
1574 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1575 * MTU (1280) and a fragment header should always be included
1576 * after a node receiving Too Big message reporting PMTU is
1577 * less than the IPv6 Minimum Link MTU.
1579 pmtu = IPV6_MIN_MTU;
1583 /* New mtu received -> path was valid.
1584 They are sent only in response to data packets,
1585 so that this nexthop apparently is reachable. --ANK
1587 dst_confirm(&rt->u.dst);
1589 /* Host route. If it is static, it would be better
1590 not to override it, but add new one, so that
1591 when cache entry will expire old pmtu
1592 would return automatically.
1594 if (rt->rt6i_flags & RTF_CACHE) {
1595 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1597 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1598 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1599 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1604 Two cases are possible:
1605 1. It is connected route. Action: COW
1606 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1608 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1609 nrt = rt6_alloc_cow(rt, daddr, saddr);
1611 nrt = rt6_alloc_clone(rt, daddr);
1614 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1616 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1618 /* According to RFC 1981, detecting PMTU increase shouldn't be
1619 * happened within 5 mins, the recommended timer is 10 mins.
1620 * Here this route expiration time is set to ip6_rt_mtu_expires
1621 * which is 10 mins. After 10 mins the decreased pmtu is expired
1622 * and detecting PMTU increase will be automatically happened.
1624 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1625 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1630 dst_release(&rt->u.dst);
1634 * Misc support functions
1637 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1639 struct net *net = dev_net(ort->rt6i_dev);
1640 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1643 rt->u.dst.input = ort->u.dst.input;
1644 rt->u.dst.output = ort->u.dst.output;
1646 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1647 rt->u.dst.error = ort->u.dst.error;
1648 rt->u.dst.dev = ort->u.dst.dev;
1650 dev_hold(rt->u.dst.dev);
1651 rt->rt6i_idev = ort->rt6i_idev;
1653 in6_dev_hold(rt->rt6i_idev);
1654 rt->u.dst.lastuse = jiffies;
1655 rt->rt6i_expires = 0;
1657 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1658 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1659 rt->rt6i_metric = 0;
1661 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1662 #ifdef CONFIG_IPV6_SUBTREES
1663 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1665 rt->rt6i_table = ort->rt6i_table;
1670 #ifdef CONFIG_IPV6_ROUTE_INFO
1671 static struct rt6_info *rt6_get_route_info(struct net *net,
1672 struct in6_addr *prefix, int prefixlen,
1673 struct in6_addr *gwaddr, int ifindex)
1675 struct fib6_node *fn;
1676 struct rt6_info *rt = NULL;
1677 struct fib6_table *table;
1679 table = fib6_get_table(net, RT6_TABLE_INFO);
1683 write_lock_bh(&table->tb6_lock);
1684 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1688 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1689 if (rt->rt6i_dev->ifindex != ifindex)
1691 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1693 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1695 dst_hold(&rt->u.dst);
1699 write_unlock_bh(&table->tb6_lock);
1703 static struct rt6_info *rt6_add_route_info(struct net *net,
1704 struct in6_addr *prefix, int prefixlen,
1705 struct in6_addr *gwaddr, int ifindex,
1708 struct fib6_config cfg = {
1709 .fc_table = RT6_TABLE_INFO,
1710 .fc_metric = IP6_RT_PRIO_USER,
1711 .fc_ifindex = ifindex,
1712 .fc_dst_len = prefixlen,
1713 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1714 RTF_UP | RTF_PREF(pref),
1716 .fc_nlinfo.nlh = NULL,
1717 .fc_nlinfo.nl_net = net,
1720 ipv6_addr_copy(&cfg.fc_dst, prefix);
1721 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1723 /* We should treat it as a default route if prefix length is 0. */
1725 cfg.fc_flags |= RTF_DEFAULT;
1727 ip6_route_add(&cfg);
1729 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1733 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1735 struct rt6_info *rt;
1736 struct fib6_table *table;
1738 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1742 write_lock_bh(&table->tb6_lock);
1743 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1744 if (dev == rt->rt6i_dev &&
1745 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1746 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1750 dst_hold(&rt->u.dst);
1751 write_unlock_bh(&table->tb6_lock);
1755 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1756 struct net_device *dev,
1759 struct fib6_config cfg = {
1760 .fc_table = RT6_TABLE_DFLT,
1761 .fc_metric = IP6_RT_PRIO_USER,
1762 .fc_ifindex = dev->ifindex,
1763 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1764 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1766 .fc_nlinfo.nlh = NULL,
1767 .fc_nlinfo.nl_net = dev_net(dev),
1770 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1772 ip6_route_add(&cfg);
1774 return rt6_get_dflt_router(gwaddr, dev);
1777 void rt6_purge_dflt_routers(struct net *net)
1779 struct rt6_info *rt;
1780 struct fib6_table *table;
1782 /* NOTE: Keep consistent with rt6_get_dflt_router */
1783 table = fib6_get_table(net, RT6_TABLE_DFLT);
1788 read_lock_bh(&table->tb6_lock);
1789 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1790 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1791 dst_hold(&rt->u.dst);
1792 read_unlock_bh(&table->tb6_lock);
1797 read_unlock_bh(&table->tb6_lock);
1800 static void rtmsg_to_fib6_config(struct net *net,
1801 struct in6_rtmsg *rtmsg,
1802 struct fib6_config *cfg)
1804 memset(cfg, 0, sizeof(*cfg));
1806 cfg->fc_table = RT6_TABLE_MAIN;
1807 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1808 cfg->fc_metric = rtmsg->rtmsg_metric;
1809 cfg->fc_expires = rtmsg->rtmsg_info;
1810 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1811 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1812 cfg->fc_flags = rtmsg->rtmsg_flags;
1814 cfg->fc_nlinfo.nl_net = net;
1816 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1817 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1818 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1821 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1823 struct fib6_config cfg;
1824 struct in6_rtmsg rtmsg;
1828 case SIOCADDRT: /* Add a route */
1829 case SIOCDELRT: /* Delete a route */
1830 if (!capable(CAP_NET_ADMIN))
1832 err = copy_from_user(&rtmsg, arg,
1833 sizeof(struct in6_rtmsg));
1837 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1842 err = ip6_route_add(&cfg);
1845 err = ip6_route_del(&cfg);
1859 * Drop the packet on the floor
1862 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
1865 struct dst_entry *dst = skb_dst(skb);
1866 switch (ipstats_mib_noroutes) {
1867 case IPSTATS_MIB_INNOROUTES:
1868 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1869 if (type == IPV6_ADDR_ANY) {
1870 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1871 IPSTATS_MIB_INADDRERRORS);
1875 case IPSTATS_MIB_OUTNOROUTES:
1876 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1877 ipstats_mib_noroutes);
1880 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
1885 static int ip6_pkt_discard(struct sk_buff *skb)
1887 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1890 static int ip6_pkt_discard_out(struct sk_buff *skb)
1892 skb->dev = skb_dst(skb)->dev;
1893 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1896 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1898 static int ip6_pkt_prohibit(struct sk_buff *skb)
1900 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1903 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1905 skb->dev = skb_dst(skb)->dev;
1906 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1912 * Allocate a dst for local (unicast / anycast) address.
1915 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1916 const struct in6_addr *addr,
1919 struct net *net = dev_net(idev->dev);
1920 struct rt6_info *rt = ip6_dst_alloc(&net->ipv6.ip6_dst_ops);
1921 struct neighbour *neigh;
1924 return ERR_PTR(-ENOMEM);
1926 dev_hold(net->loopback_dev);
1929 rt->u.dst.flags = DST_HOST;
1930 rt->u.dst.input = ip6_input;
1931 rt->u.dst.output = ip6_output;
1932 rt->rt6i_dev = net->loopback_dev;
1933 rt->rt6i_idev = idev;
1934 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1935 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1936 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1937 rt->u.dst.obsolete = -1;
1939 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1941 rt->rt6i_flags |= RTF_ANYCAST;
1943 rt->rt6i_flags |= RTF_LOCAL;
1944 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1945 if (IS_ERR(neigh)) {
1946 dst_free(&rt->u.dst);
1948 /* We are casting this because that is the return
1949 * value type. But an errno encoded pointer is the
1950 * same regardless of the underlying pointer type,
1951 * and that's what we are returning. So this is OK.
1953 return (struct rt6_info *) neigh;
1955 rt->rt6i_nexthop = neigh;
1957 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1958 rt->rt6i_dst.plen = 128;
1959 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1961 atomic_set(&rt->u.dst.__refcnt, 1);
1966 struct arg_dev_net {
1967 struct net_device *dev;
1971 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1973 struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1974 struct net *net = ((struct arg_dev_net *)arg)->net;
1976 if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1977 rt != net->ipv6.ip6_null_entry) {
1978 RT6_TRACE("deleted by ifdown %p\n", rt);
1984 void rt6_ifdown(struct net *net, struct net_device *dev)
1986 struct arg_dev_net adn = {
1991 fib6_clean_all(net, fib6_ifdown, 0, &adn);
1992 icmp6_clean_all(fib6_ifdown, &adn);
1995 struct rt6_mtu_change_arg
1997 struct net_device *dev;
2001 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2003 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2004 struct inet6_dev *idev;
2005 struct net *net = dev_net(arg->dev);
2007 /* In IPv6 pmtu discovery is not optional,
2008 so that RTAX_MTU lock cannot disable it.
2009 We still use this lock to block changes
2010 caused by addrconf/ndisc.
2013 idev = __in6_dev_get(arg->dev);
2017 /* For administrative MTU increase, there is no way to discover
2018 IPv6 PMTU increase, so PMTU increase should be updated here.
2019 Since RFC 1981 doesn't include administrative MTU increase
2020 update PMTU increase is a MUST. (i.e. jumbo frame)
2023 If new MTU is less than route PMTU, this new MTU will be the
2024 lowest MTU in the path, update the route PMTU to reflect PMTU
2025 decreases; if new MTU is greater than route PMTU, and the
2026 old MTU is the lowest MTU in the path, update the route PMTU
2027 to reflect the increase. In this case if the other nodes' MTU
2028 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2031 if (rt->rt6i_dev == arg->dev &&
2032 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
2033 (dst_mtu(&rt->u.dst) >= arg->mtu ||
2034 (dst_mtu(&rt->u.dst) < arg->mtu &&
2035 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
2036 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
2037 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
2042 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2044 struct rt6_mtu_change_arg arg = {
2049 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2052 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2053 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2054 [RTA_OIF] = { .type = NLA_U32 },
2055 [RTA_IIF] = { .type = NLA_U32 },
2056 [RTA_PRIORITY] = { .type = NLA_U32 },
2057 [RTA_METRICS] = { .type = NLA_NESTED },
2060 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2061 struct fib6_config *cfg)
2064 struct nlattr *tb[RTA_MAX+1];
2067 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2072 rtm = nlmsg_data(nlh);
2073 memset(cfg, 0, sizeof(*cfg));
2075 cfg->fc_table = rtm->rtm_table;
2076 cfg->fc_dst_len = rtm->rtm_dst_len;
2077 cfg->fc_src_len = rtm->rtm_src_len;
2078 cfg->fc_flags = RTF_UP;
2079 cfg->fc_protocol = rtm->rtm_protocol;
2081 if (rtm->rtm_type == RTN_UNREACHABLE)
2082 cfg->fc_flags |= RTF_REJECT;
2084 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2085 cfg->fc_nlinfo.nlh = nlh;
2086 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2088 if (tb[RTA_GATEWAY]) {
2089 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2090 cfg->fc_flags |= RTF_GATEWAY;
2094 int plen = (rtm->rtm_dst_len + 7) >> 3;
2096 if (nla_len(tb[RTA_DST]) < plen)
2099 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2103 int plen = (rtm->rtm_src_len + 7) >> 3;
2105 if (nla_len(tb[RTA_SRC]) < plen)
2108 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2112 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2114 if (tb[RTA_PRIORITY])
2115 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2117 if (tb[RTA_METRICS]) {
2118 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2119 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2123 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2130 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2132 struct fib6_config cfg;
2135 err = rtm_to_fib6_config(skb, nlh, &cfg);
2139 return ip6_route_del(&cfg);
2142 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2144 struct fib6_config cfg;
2147 err = rtm_to_fib6_config(skb, nlh, &cfg);
2151 return ip6_route_add(&cfg);
2154 static inline size_t rt6_nlmsg_size(void)
2156 return NLMSG_ALIGN(sizeof(struct rtmsg))
2157 + nla_total_size(16) /* RTA_SRC */
2158 + nla_total_size(16) /* RTA_DST */
2159 + nla_total_size(16) /* RTA_GATEWAY */
2160 + nla_total_size(16) /* RTA_PREFSRC */
2161 + nla_total_size(4) /* RTA_TABLE */
2162 + nla_total_size(4) /* RTA_IIF */
2163 + nla_total_size(4) /* RTA_OIF */
2164 + nla_total_size(4) /* RTA_PRIORITY */
2165 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2166 + nla_total_size(sizeof(struct rta_cacheinfo));
2169 static int rt6_fill_node(struct net *net,
2170 struct sk_buff *skb, struct rt6_info *rt,
2171 struct in6_addr *dst, struct in6_addr *src,
2172 int iif, int type, u32 pid, u32 seq,
2173 int prefix, int nowait, unsigned int flags)
2176 struct nlmsghdr *nlh;
2180 if (prefix) { /* user wants prefix routes only */
2181 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2182 /* success since this is not a prefix route */
2187 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2191 rtm = nlmsg_data(nlh);
2192 rtm->rtm_family = AF_INET6;
2193 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2194 rtm->rtm_src_len = rt->rt6i_src.plen;
2197 table = rt->rt6i_table->tb6_id;
2199 table = RT6_TABLE_UNSPEC;
2200 rtm->rtm_table = table;
2201 NLA_PUT_U32(skb, RTA_TABLE, table);
2202 if (rt->rt6i_flags&RTF_REJECT)
2203 rtm->rtm_type = RTN_UNREACHABLE;
2204 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2205 rtm->rtm_type = RTN_LOCAL;
2207 rtm->rtm_type = RTN_UNICAST;
2209 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2210 rtm->rtm_protocol = rt->rt6i_protocol;
2211 if (rt->rt6i_flags&RTF_DYNAMIC)
2212 rtm->rtm_protocol = RTPROT_REDIRECT;
2213 else if (rt->rt6i_flags & RTF_ADDRCONF)
2214 rtm->rtm_protocol = RTPROT_KERNEL;
2215 else if (rt->rt6i_flags&RTF_DEFAULT)
2216 rtm->rtm_protocol = RTPROT_RA;
2218 if (rt->rt6i_flags&RTF_CACHE)
2219 rtm->rtm_flags |= RTM_F_CLONED;
2222 NLA_PUT(skb, RTA_DST, 16, dst);
2223 rtm->rtm_dst_len = 128;
2224 } else if (rtm->rtm_dst_len)
2225 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2226 #ifdef CONFIG_IPV6_SUBTREES
2228 NLA_PUT(skb, RTA_SRC, 16, src);
2229 rtm->rtm_src_len = 128;
2230 } else if (rtm->rtm_src_len)
2231 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2234 #ifdef CONFIG_IPV6_MROUTE
2235 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2236 int err = ip6mr_get_route(net, skb, rtm, nowait);
2241 goto nla_put_failure;
2243 if (err == -EMSGSIZE)
2244 goto nla_put_failure;
2249 NLA_PUT_U32(skb, RTA_IIF, iif);
2251 struct inet6_dev *idev = ip6_dst_idev(&rt->u.dst);
2252 struct in6_addr saddr_buf;
2253 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2254 dst, 0, &saddr_buf) == 0)
2255 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2258 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2259 goto nla_put_failure;
2261 if (rt->u.dst.neighbour)
2262 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2265 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2267 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2269 if (!(rt->rt6i_flags & RTF_EXPIRES))
2271 else if (rt->rt6i_expires - jiffies < INT_MAX)
2272 expires = rt->rt6i_expires - jiffies;
2276 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2277 expires, rt->u.dst.error) < 0)
2278 goto nla_put_failure;
2280 return nlmsg_end(skb, nlh);
2283 nlmsg_cancel(skb, nlh);
2287 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2289 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2292 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2293 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2294 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2298 return rt6_fill_node(arg->net,
2299 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2300 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2301 prefix, 0, NLM_F_MULTI);
2304 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2306 struct net *net = sock_net(in_skb->sk);
2307 struct nlattr *tb[RTA_MAX+1];
2308 struct rt6_info *rt;
2309 struct sk_buff *skb;
2314 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2319 memset(&fl, 0, sizeof(fl));
2322 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2325 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2329 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2332 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2336 iif = nla_get_u32(tb[RTA_IIF]);
2339 fl.oif = nla_get_u32(tb[RTA_OIF]);
2342 struct net_device *dev;
2343 dev = __dev_get_by_index(net, iif);
2350 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2356 /* Reserve room for dummy headers, this skb can pass
2357 through good chunk of routing engine.
2359 skb_reset_mac_header(skb);
2360 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2362 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2363 skb_dst_set(skb, &rt->u.dst);
2365 err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2366 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2367 nlh->nlmsg_seq, 0, 0, 0);
2373 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2378 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2380 struct sk_buff *skb;
2381 struct net *net = info->nl_net;
2386 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2388 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2392 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2393 event, info->pid, seq, 0, 0, 0);
2395 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2396 WARN_ON(err == -EMSGSIZE);
2400 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2401 info->nlh, gfp_any());
2405 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2408 static int ip6_route_dev_notify(struct notifier_block *this,
2409 unsigned long event, void *data)
2411 struct net_device *dev = (struct net_device *)data;
2412 struct net *net = dev_net(dev);
2414 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2415 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2416 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2417 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2418 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2419 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2420 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2421 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2432 #ifdef CONFIG_PROC_FS
2434 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2445 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2447 struct seq_file *m = p_arg;
2449 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2451 #ifdef CONFIG_IPV6_SUBTREES
2452 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2454 seq_puts(m, "00000000000000000000000000000000 00 ");
2457 if (rt->rt6i_nexthop) {
2458 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2460 seq_puts(m, "00000000000000000000000000000000");
2462 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2463 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2464 rt->u.dst.__use, rt->rt6i_flags,
2465 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2469 static int ipv6_route_show(struct seq_file *m, void *v)
2471 struct net *net = (struct net *)m->private;
2472 fib6_clean_all(net, rt6_info_route, 0, m);
2476 static int ipv6_route_open(struct inode *inode, struct file *file)
2478 return single_open_net(inode, file, ipv6_route_show);
2481 static const struct file_operations ipv6_route_proc_fops = {
2482 .owner = THIS_MODULE,
2483 .open = ipv6_route_open,
2485 .llseek = seq_lseek,
2486 .release = single_release_net,
2489 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2491 struct net *net = (struct net *)seq->private;
2492 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2493 net->ipv6.rt6_stats->fib_nodes,
2494 net->ipv6.rt6_stats->fib_route_nodes,
2495 net->ipv6.rt6_stats->fib_rt_alloc,
2496 net->ipv6.rt6_stats->fib_rt_entries,
2497 net->ipv6.rt6_stats->fib_rt_cache,
2498 atomic_read(&net->ipv6.ip6_dst_ops.entries),
2499 net->ipv6.rt6_stats->fib_discarded_routes);
2504 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2506 return single_open_net(inode, file, rt6_stats_seq_show);
2509 static const struct file_operations rt6_stats_seq_fops = {
2510 .owner = THIS_MODULE,
2511 .open = rt6_stats_seq_open,
2513 .llseek = seq_lseek,
2514 .release = single_release_net,
2516 #endif /* CONFIG_PROC_FS */
2518 #ifdef CONFIG_SYSCTL
2521 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write,
2522 void __user *buffer, size_t *lenp, loff_t *ppos)
2524 struct net *net = current->nsproxy->net_ns;
2525 int delay = net->ipv6.sysctl.flush_delay;
2527 proc_dointvec(ctl, write, buffer, lenp, ppos);
2528 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2534 ctl_table ipv6_route_table_template[] = {
2536 .procname = "flush",
2537 .data = &init_net.ipv6.sysctl.flush_delay,
2538 .maxlen = sizeof(int),
2540 .proc_handler = ipv6_sysctl_rtcache_flush
2543 .procname = "gc_thresh",
2544 .data = &ip6_dst_ops_template.gc_thresh,
2545 .maxlen = sizeof(int),
2547 .proc_handler = proc_dointvec,
2550 .procname = "max_size",
2551 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2552 .maxlen = sizeof(int),
2554 .proc_handler = proc_dointvec,
2557 .procname = "gc_min_interval",
2558 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2559 .maxlen = sizeof(int),
2561 .proc_handler = proc_dointvec_jiffies,
2564 .procname = "gc_timeout",
2565 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2566 .maxlen = sizeof(int),
2568 .proc_handler = proc_dointvec_jiffies,
2571 .procname = "gc_interval",
2572 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2573 .maxlen = sizeof(int),
2575 .proc_handler = proc_dointvec_jiffies,
2578 .procname = "gc_elasticity",
2579 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2580 .maxlen = sizeof(int),
2582 .proc_handler = proc_dointvec_jiffies,
2585 .procname = "mtu_expires",
2586 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2587 .maxlen = sizeof(int),
2589 .proc_handler = proc_dointvec_jiffies,
2592 .procname = "min_adv_mss",
2593 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2594 .maxlen = sizeof(int),
2596 .proc_handler = proc_dointvec_jiffies,
2599 .procname = "gc_min_interval_ms",
2600 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2601 .maxlen = sizeof(int),
2603 .proc_handler = proc_dointvec_ms_jiffies,
2608 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
2610 struct ctl_table *table;
2612 table = kmemdup(ipv6_route_table_template,
2613 sizeof(ipv6_route_table_template),
2617 table[0].data = &net->ipv6.sysctl.flush_delay;
2618 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
2619 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2620 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2621 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2622 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2623 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2624 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2625 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2626 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2633 static int __net_init ip6_route_net_init(struct net *net)
2637 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
2638 sizeof(net->ipv6.ip6_dst_ops));
2640 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2641 sizeof(*net->ipv6.ip6_null_entry),
2643 if (!net->ipv6.ip6_null_entry)
2644 goto out_ip6_dst_ops;
2645 net->ipv6.ip6_null_entry->u.dst.path =
2646 (struct dst_entry *)net->ipv6.ip6_null_entry;
2647 net->ipv6.ip6_null_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2649 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2650 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2651 sizeof(*net->ipv6.ip6_prohibit_entry),
2653 if (!net->ipv6.ip6_prohibit_entry)
2654 goto out_ip6_null_entry;
2655 net->ipv6.ip6_prohibit_entry->u.dst.path =
2656 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2657 net->ipv6.ip6_prohibit_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2659 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2660 sizeof(*net->ipv6.ip6_blk_hole_entry),
2662 if (!net->ipv6.ip6_blk_hole_entry)
2663 goto out_ip6_prohibit_entry;
2664 net->ipv6.ip6_blk_hole_entry->u.dst.path =
2665 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2666 net->ipv6.ip6_blk_hole_entry->u.dst.ops = &net->ipv6.ip6_dst_ops;
2669 net->ipv6.sysctl.flush_delay = 0;
2670 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2671 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2672 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2673 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2674 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2675 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2676 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2678 #ifdef CONFIG_PROC_FS
2679 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2680 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2682 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2688 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2689 out_ip6_prohibit_entry:
2690 kfree(net->ipv6.ip6_prohibit_entry);
2692 kfree(net->ipv6.ip6_null_entry);
2698 static void __net_exit ip6_route_net_exit(struct net *net)
2700 #ifdef CONFIG_PROC_FS
2701 proc_net_remove(net, "ipv6_route");
2702 proc_net_remove(net, "rt6_stats");
2704 kfree(net->ipv6.ip6_null_entry);
2705 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2706 kfree(net->ipv6.ip6_prohibit_entry);
2707 kfree(net->ipv6.ip6_blk_hole_entry);
2711 static struct pernet_operations ip6_route_net_ops = {
2712 .init = ip6_route_net_init,
2713 .exit = ip6_route_net_exit,
2716 static struct notifier_block ip6_route_dev_notifier = {
2717 .notifier_call = ip6_route_dev_notify,
2721 int __init ip6_route_init(void)
2726 ip6_dst_ops_template.kmem_cachep =
2727 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2728 SLAB_HWCACHE_ALIGN, NULL);
2729 if (!ip6_dst_ops_template.kmem_cachep)
2732 ret = register_pernet_subsys(&ip6_route_net_ops);
2734 goto out_kmem_cache;
2736 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2738 /* Registering of the loopback is done before this portion of code,
2739 * the loopback reference in rt6_info will not be taken, do it
2740 * manually for init_net */
2741 init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2742 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2743 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2744 init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2745 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2746 init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2747 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2751 goto out_register_subsys;
2757 ret = fib6_rules_init();
2762 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2763 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2764 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2765 goto fib6_rules_init;
2767 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2769 goto fib6_rules_init;
2775 fib6_rules_cleanup();
2780 out_register_subsys:
2781 unregister_pernet_subsys(&ip6_route_net_ops);
2783 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2787 void ip6_route_cleanup(void)
2789 unregister_netdevice_notifier(&ip6_route_dev_notifier);
2790 fib6_rules_cleanup();
2793 unregister_pernet_subsys(&ip6_route_net_ops);
2794 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);