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)
484 pref = ICMPV6_ROUTER_PREF_MEDIUM;
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 unsigned int prefs = inet6_sk(sk)->srcprefs;
824 if (prefs & IPV6_PREFER_SRC_TMP)
825 flags |= RT6_LOOKUP_F_SRCPREF_TMP;
826 if (prefs & IPV6_PREFER_SRC_PUBLIC)
827 flags |= RT6_LOOKUP_F_SRCPREF_PUBLIC;
828 if (prefs & IPV6_PREFER_SRC_COA)
829 flags |= RT6_LOOKUP_F_SRCPREF_COA;
832 return fib6_rule_lookup(net, fl, flags, ip6_pol_route_output);
835 EXPORT_SYMBOL(ip6_route_output);
837 int ip6_dst_blackhole(struct sock *sk, struct dst_entry **dstp, struct flowi *fl)
839 struct rt6_info *ort = (struct rt6_info *) *dstp;
840 struct rt6_info *rt = (struct rt6_info *)
841 dst_alloc(&ip6_dst_blackhole_ops);
842 struct dst_entry *new = NULL;
847 atomic_set(&new->__refcnt, 1);
849 new->input = dst_discard;
850 new->output = dst_discard;
852 memcpy(new->metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
853 new->dev = ort->u.dst.dev;
856 rt->rt6i_idev = ort->rt6i_idev;
858 in6_dev_hold(rt->rt6i_idev);
859 rt->rt6i_expires = 0;
861 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
862 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
865 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
866 #ifdef CONFIG_IPV6_SUBTREES
867 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
875 return (new ? 0 : -ENOMEM);
877 EXPORT_SYMBOL_GPL(ip6_dst_blackhole);
880 * Destination cache support functions
883 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
887 rt = (struct rt6_info *) dst;
889 if (rt && rt->rt6i_node && (rt->rt6i_node->fn_sernum == cookie))
895 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
897 struct rt6_info *rt = (struct rt6_info *) dst;
900 if (rt->rt6i_flags & RTF_CACHE)
908 static void ip6_link_failure(struct sk_buff *skb)
912 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0, skb->dev);
914 rt = (struct rt6_info *) skb_dst(skb);
916 if (rt->rt6i_flags&RTF_CACHE) {
917 dst_set_expires(&rt->u.dst, 0);
918 rt->rt6i_flags |= RTF_EXPIRES;
919 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT))
920 rt->rt6i_node->fn_sernum = -1;
924 static void ip6_rt_update_pmtu(struct dst_entry *dst, u32 mtu)
926 struct rt6_info *rt6 = (struct rt6_info*)dst;
928 if (mtu < dst_mtu(dst) && rt6->rt6i_dst.plen == 128) {
929 rt6->rt6i_flags |= RTF_MODIFIED;
930 if (mtu < IPV6_MIN_MTU) {
932 dst->metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
934 dst->metrics[RTAX_MTU-1] = mtu;
935 call_netevent_notifiers(NETEVENT_PMTU_UPDATE, dst);
939 static int ipv6_get_mtu(struct net_device *dev);
941 static inline unsigned int ipv6_advmss(struct net *net, unsigned int mtu)
943 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
945 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
946 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
949 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
950 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
951 * IPV6_MAXPLEN is also valid and means: "any MSS,
952 * rely only on pmtu discovery"
954 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
959 static struct dst_entry *icmp6_dst_gc_list;
960 static DEFINE_SPINLOCK(icmp6_dst_lock);
962 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
963 struct neighbour *neigh,
964 const struct in6_addr *addr)
967 struct inet6_dev *idev = in6_dev_get(dev);
968 struct net *net = dev_net(dev);
970 if (unlikely(idev == NULL))
973 rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
974 if (unlikely(rt == NULL)) {
983 neigh = ndisc_get_neigh(dev, addr);
989 rt->rt6i_idev = idev;
990 rt->rt6i_nexthop = neigh;
991 atomic_set(&rt->u.dst.__refcnt, 1);
992 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = 255;
993 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
994 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
995 rt->u.dst.output = ip6_output;
997 #if 0 /* there's no chance to use these for ndisc */
998 rt->u.dst.flags = ipv6_addr_type(addr) & IPV6_ADDR_UNICAST
1001 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1002 rt->rt6i_dst.plen = 128;
1005 spin_lock_bh(&icmp6_dst_lock);
1006 rt->u.dst.next = icmp6_dst_gc_list;
1007 icmp6_dst_gc_list = &rt->u.dst;
1008 spin_unlock_bh(&icmp6_dst_lock);
1010 fib6_force_start_gc(net);
1016 int icmp6_dst_gc(void)
1018 struct dst_entry *dst, *next, **pprev;
1023 spin_lock_bh(&icmp6_dst_lock);
1024 pprev = &icmp6_dst_gc_list;
1026 while ((dst = *pprev) != NULL) {
1027 if (!atomic_read(&dst->__refcnt)) {
1036 spin_unlock_bh(&icmp6_dst_lock);
1041 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1044 struct dst_entry *dst, **pprev;
1046 spin_lock_bh(&icmp6_dst_lock);
1047 pprev = &icmp6_dst_gc_list;
1048 while ((dst = *pprev) != NULL) {
1049 struct rt6_info *rt = (struct rt6_info *) dst;
1050 if (func(rt, arg)) {
1057 spin_unlock_bh(&icmp6_dst_lock);
1060 static int ip6_dst_gc(struct dst_ops *ops)
1062 unsigned long now = jiffies;
1063 struct net *net = ops->dst_net;
1064 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1065 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1066 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1067 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1068 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1070 if (time_after(rt_last_gc + rt_min_interval, now) &&
1071 atomic_read(&ops->entries) <= rt_max_size)
1074 net->ipv6.ip6_rt_gc_expire++;
1075 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net);
1076 net->ipv6.ip6_rt_last_gc = now;
1077 if (atomic_read(&ops->entries) < ops->gc_thresh)
1078 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1080 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1081 return (atomic_read(&ops->entries) > rt_max_size);
1084 /* Clean host part of a prefix. Not necessary in radix tree,
1085 but results in cleaner routing tables.
1087 Remove it only when all the things will work!
1090 static int ipv6_get_mtu(struct net_device *dev)
1092 int mtu = IPV6_MIN_MTU;
1093 struct inet6_dev *idev;
1095 idev = in6_dev_get(dev);
1097 mtu = idev->cnf.mtu6;
1103 int ip6_dst_hoplimit(struct dst_entry *dst)
1105 int hoplimit = dst_metric(dst, RTAX_HOPLIMIT);
1107 struct net_device *dev = dst->dev;
1108 struct inet6_dev *idev = in6_dev_get(dev);
1110 hoplimit = idev->cnf.hop_limit;
1113 hoplimit = dev_net(dev)->ipv6.devconf_all->hop_limit;
1122 int ip6_route_add(struct fib6_config *cfg)
1125 struct net *net = cfg->fc_nlinfo.nl_net;
1126 struct rt6_info *rt = NULL;
1127 struct net_device *dev = NULL;
1128 struct inet6_dev *idev = NULL;
1129 struct fib6_table *table;
1132 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1134 #ifndef CONFIG_IPV6_SUBTREES
1135 if (cfg->fc_src_len)
1138 if (cfg->fc_ifindex) {
1140 dev = dev_get_by_index(net, cfg->fc_ifindex);
1143 idev = in6_dev_get(dev);
1148 if (cfg->fc_metric == 0)
1149 cfg->fc_metric = IP6_RT_PRIO_USER;
1151 table = fib6_new_table(net, cfg->fc_table);
1152 if (table == NULL) {
1157 rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1164 rt->u.dst.obsolete = -1;
1165 rt->rt6i_expires = (cfg->fc_flags & RTF_EXPIRES) ?
1166 jiffies + clock_t_to_jiffies(cfg->fc_expires) :
1169 if (cfg->fc_protocol == RTPROT_UNSPEC)
1170 cfg->fc_protocol = RTPROT_BOOT;
1171 rt->rt6i_protocol = cfg->fc_protocol;
1173 addr_type = ipv6_addr_type(&cfg->fc_dst);
1175 if (addr_type & IPV6_ADDR_MULTICAST)
1176 rt->u.dst.input = ip6_mc_input;
1178 rt->u.dst.input = ip6_forward;
1180 rt->u.dst.output = ip6_output;
1182 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1183 rt->rt6i_dst.plen = cfg->fc_dst_len;
1184 if (rt->rt6i_dst.plen == 128)
1185 rt->u.dst.flags = DST_HOST;
1187 #ifdef CONFIG_IPV6_SUBTREES
1188 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1189 rt->rt6i_src.plen = cfg->fc_src_len;
1192 rt->rt6i_metric = cfg->fc_metric;
1194 /* We cannot add true routes via loopback here,
1195 they would result in kernel looping; promote them to reject routes
1197 if ((cfg->fc_flags & RTF_REJECT) ||
1198 (dev && (dev->flags&IFF_LOOPBACK) && !(addr_type&IPV6_ADDR_LOOPBACK))) {
1199 /* hold loopback dev/idev if we haven't done so. */
1200 if (dev != net->loopback_dev) {
1205 dev = net->loopback_dev;
1207 idev = in6_dev_get(dev);
1213 rt->u.dst.output = ip6_pkt_discard_out;
1214 rt->u.dst.input = ip6_pkt_discard;
1215 rt->u.dst.error = -ENETUNREACH;
1216 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1220 if (cfg->fc_flags & RTF_GATEWAY) {
1221 struct in6_addr *gw_addr;
1224 gw_addr = &cfg->fc_gateway;
1225 ipv6_addr_copy(&rt->rt6i_gateway, gw_addr);
1226 gwa_type = ipv6_addr_type(gw_addr);
1228 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1229 struct rt6_info *grt;
1231 /* IPv6 strictly inhibits using not link-local
1232 addresses as nexthop address.
1233 Otherwise, router will not able to send redirects.
1234 It is very good, but in some (rare!) circumstances
1235 (SIT, PtP, NBMA NOARP links) it is handy to allow
1236 some exceptions. --ANK
1239 if (!(gwa_type&IPV6_ADDR_UNICAST))
1242 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1244 err = -EHOSTUNREACH;
1248 if (dev != grt->rt6i_dev) {
1249 dst_release(&grt->u.dst);
1253 dev = grt->rt6i_dev;
1254 idev = grt->rt6i_idev;
1256 in6_dev_hold(grt->rt6i_idev);
1258 if (!(grt->rt6i_flags&RTF_GATEWAY))
1260 dst_release(&grt->u.dst);
1266 if (dev == NULL || (dev->flags&IFF_LOOPBACK))
1274 if (cfg->fc_flags & (RTF_GATEWAY | RTF_NONEXTHOP)) {
1275 rt->rt6i_nexthop = __neigh_lookup_errno(&nd_tbl, &rt->rt6i_gateway, dev);
1276 if (IS_ERR(rt->rt6i_nexthop)) {
1277 err = PTR_ERR(rt->rt6i_nexthop);
1278 rt->rt6i_nexthop = NULL;
1283 rt->rt6i_flags = cfg->fc_flags;
1290 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1291 int type = nla_type(nla);
1294 if (type > RTAX_MAX) {
1299 rt->u.dst.metrics[type - 1] = nla_get_u32(nla);
1304 if (dst_metric(&rt->u.dst, RTAX_HOPLIMIT) == 0)
1305 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1306 if (!dst_mtu(&rt->u.dst))
1307 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(dev);
1308 if (!dst_metric(&rt->u.dst, RTAX_ADVMSS))
1309 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1310 rt->u.dst.dev = dev;
1311 rt->rt6i_idev = idev;
1312 rt->rt6i_table = table;
1314 cfg->fc_nlinfo.nl_net = dev_net(dev);
1316 return __ip6_ins_rt(rt, &cfg->fc_nlinfo);
1324 dst_free(&rt->u.dst);
1328 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
1331 struct fib6_table *table;
1332 struct net *net = dev_net(rt->rt6i_dev);
1334 if (rt == net->ipv6.ip6_null_entry)
1337 table = rt->rt6i_table;
1338 write_lock_bh(&table->tb6_lock);
1340 err = fib6_del(rt, info);
1341 dst_release(&rt->u.dst);
1343 write_unlock_bh(&table->tb6_lock);
1348 int ip6_del_rt(struct rt6_info *rt)
1350 struct nl_info info = {
1351 .nl_net = dev_net(rt->rt6i_dev),
1353 return __ip6_del_rt(rt, &info);
1356 static int ip6_route_del(struct fib6_config *cfg)
1358 struct fib6_table *table;
1359 struct fib6_node *fn;
1360 struct rt6_info *rt;
1363 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
1367 read_lock_bh(&table->tb6_lock);
1369 fn = fib6_locate(&table->tb6_root,
1370 &cfg->fc_dst, cfg->fc_dst_len,
1371 &cfg->fc_src, cfg->fc_src_len);
1374 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1375 if (cfg->fc_ifindex &&
1376 (rt->rt6i_dev == NULL ||
1377 rt->rt6i_dev->ifindex != cfg->fc_ifindex))
1379 if (cfg->fc_flags & RTF_GATEWAY &&
1380 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
1382 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
1384 dst_hold(&rt->u.dst);
1385 read_unlock_bh(&table->tb6_lock);
1387 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
1390 read_unlock_bh(&table->tb6_lock);
1398 struct ip6rd_flowi {
1400 struct in6_addr gateway;
1403 static struct rt6_info *__ip6_route_redirect(struct net *net,
1404 struct fib6_table *table,
1408 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl;
1409 struct rt6_info *rt;
1410 struct fib6_node *fn;
1413 * Get the "current" route for this destination and
1414 * check if the redirect has come from approriate router.
1416 * RFC 2461 specifies that redirects should only be
1417 * accepted if they come from the nexthop to the target.
1418 * Due to the way the routes are chosen, this notion
1419 * is a bit fuzzy and one might need to check all possible
1423 read_lock_bh(&table->tb6_lock);
1424 fn = fib6_lookup(&table->tb6_root, &fl->fl6_dst, &fl->fl6_src);
1426 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1428 * Current route is on-link; redirect is always invalid.
1430 * Seems, previous statement is not true. It could
1431 * be node, which looks for us as on-link (f.e. proxy ndisc)
1432 * But then router serving it might decide, that we should
1433 * know truth 8)8) --ANK (980726).
1435 if (rt6_check_expired(rt))
1437 if (!(rt->rt6i_flags & RTF_GATEWAY))
1439 if (fl->oif != rt->rt6i_dev->ifindex)
1441 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1447 rt = net->ipv6.ip6_null_entry;
1448 BACKTRACK(net, &fl->fl6_src);
1450 dst_hold(&rt->u.dst);
1452 read_unlock_bh(&table->tb6_lock);
1457 static struct rt6_info *ip6_route_redirect(struct in6_addr *dest,
1458 struct in6_addr *src,
1459 struct in6_addr *gateway,
1460 struct net_device *dev)
1462 int flags = RT6_LOOKUP_F_HAS_SADDR;
1463 struct net *net = dev_net(dev);
1464 struct ip6rd_flowi rdfl = {
1466 .oif = dev->ifindex,
1474 .gateway = *gateway,
1477 if (rt6_need_strict(dest))
1478 flags |= RT6_LOOKUP_F_IFACE;
1480 return (struct rt6_info *)fib6_rule_lookup(net, (struct flowi *)&rdfl,
1481 flags, __ip6_route_redirect);
1484 void rt6_redirect(struct in6_addr *dest, struct in6_addr *src,
1485 struct in6_addr *saddr,
1486 struct neighbour *neigh, u8 *lladdr, int on_link)
1488 struct rt6_info *rt, *nrt = NULL;
1489 struct netevent_redirect netevent;
1490 struct net *net = dev_net(neigh->dev);
1492 rt = ip6_route_redirect(dest, src, saddr, neigh->dev);
1494 if (rt == net->ipv6.ip6_null_entry) {
1495 if (net_ratelimit())
1496 printk(KERN_DEBUG "rt6_redirect: source isn't a valid nexthop "
1497 "for redirect target\n");
1502 * We have finally decided to accept it.
1505 neigh_update(neigh, lladdr, NUD_STALE,
1506 NEIGH_UPDATE_F_WEAK_OVERRIDE|
1507 NEIGH_UPDATE_F_OVERRIDE|
1508 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
1509 NEIGH_UPDATE_F_ISROUTER))
1513 * Redirect received -> path was valid.
1514 * Look, redirects are sent only in response to data packets,
1515 * so that this nexthop apparently is reachable. --ANK
1517 dst_confirm(&rt->u.dst);
1519 /* Duplicate redirect: silently ignore. */
1520 if (neigh == rt->u.dst.neighbour)
1523 nrt = ip6_rt_copy(rt);
1527 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
1529 nrt->rt6i_flags &= ~RTF_GATEWAY;
1531 ipv6_addr_copy(&nrt->rt6i_dst.addr, dest);
1532 nrt->rt6i_dst.plen = 128;
1533 nrt->u.dst.flags |= DST_HOST;
1535 ipv6_addr_copy(&nrt->rt6i_gateway, (struct in6_addr*)neigh->primary_key);
1536 nrt->rt6i_nexthop = neigh_clone(neigh);
1537 /* Reset pmtu, it may be better */
1538 nrt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(neigh->dev);
1539 nrt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(dev_net(neigh->dev),
1540 dst_mtu(&nrt->u.dst));
1542 if (ip6_ins_rt(nrt))
1545 netevent.old = &rt->u.dst;
1546 netevent.new = &nrt->u.dst;
1547 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
1549 if (rt->rt6i_flags&RTF_CACHE) {
1555 dst_release(&rt->u.dst);
1560 * Handle ICMP "packet too big" messages
1561 * i.e. Path MTU discovery
1564 void rt6_pmtu_discovery(struct in6_addr *daddr, struct in6_addr *saddr,
1565 struct net_device *dev, u32 pmtu)
1567 struct rt6_info *rt, *nrt;
1568 struct net *net = dev_net(dev);
1571 rt = rt6_lookup(net, daddr, saddr, dev->ifindex, 0);
1575 if (pmtu >= dst_mtu(&rt->u.dst))
1578 if (pmtu < IPV6_MIN_MTU) {
1580 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1581 * MTU (1280) and a fragment header should always be included
1582 * after a node receiving Too Big message reporting PMTU is
1583 * less than the IPv6 Minimum Link MTU.
1585 pmtu = IPV6_MIN_MTU;
1589 /* New mtu received -> path was valid.
1590 They are sent only in response to data packets,
1591 so that this nexthop apparently is reachable. --ANK
1593 dst_confirm(&rt->u.dst);
1595 /* Host route. If it is static, it would be better
1596 not to override it, but add new one, so that
1597 when cache entry will expire old pmtu
1598 would return automatically.
1600 if (rt->rt6i_flags & RTF_CACHE) {
1601 rt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1603 rt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1604 dst_set_expires(&rt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1605 rt->rt6i_flags |= RTF_MODIFIED|RTF_EXPIRES;
1610 Two cases are possible:
1611 1. It is connected route. Action: COW
1612 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1614 if (!rt->rt6i_nexthop && !(rt->rt6i_flags & RTF_NONEXTHOP))
1615 nrt = rt6_alloc_cow(rt, daddr, saddr);
1617 nrt = rt6_alloc_clone(rt, daddr);
1620 nrt->u.dst.metrics[RTAX_MTU-1] = pmtu;
1622 nrt->u.dst.metrics[RTAX_FEATURES-1] |= RTAX_FEATURE_ALLFRAG;
1624 /* According to RFC 1981, detecting PMTU increase shouldn't be
1625 * happened within 5 mins, the recommended timer is 10 mins.
1626 * Here this route expiration time is set to ip6_rt_mtu_expires
1627 * which is 10 mins. After 10 mins the decreased pmtu is expired
1628 * and detecting PMTU increase will be automatically happened.
1630 dst_set_expires(&nrt->u.dst, net->ipv6.sysctl.ip6_rt_mtu_expires);
1631 nrt->rt6i_flags |= RTF_DYNAMIC|RTF_EXPIRES;
1636 dst_release(&rt->u.dst);
1640 * Misc support functions
1643 static struct rt6_info * ip6_rt_copy(struct rt6_info *ort)
1645 struct net *net = dev_net(ort->rt6i_dev);
1646 struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1649 rt->u.dst.input = ort->u.dst.input;
1650 rt->u.dst.output = ort->u.dst.output;
1652 memcpy(rt->u.dst.metrics, ort->u.dst.metrics, RTAX_MAX*sizeof(u32));
1653 rt->u.dst.error = ort->u.dst.error;
1654 rt->u.dst.dev = ort->u.dst.dev;
1656 dev_hold(rt->u.dst.dev);
1657 rt->rt6i_idev = ort->rt6i_idev;
1659 in6_dev_hold(rt->rt6i_idev);
1660 rt->u.dst.lastuse = jiffies;
1661 rt->rt6i_expires = 0;
1663 ipv6_addr_copy(&rt->rt6i_gateway, &ort->rt6i_gateway);
1664 rt->rt6i_flags = ort->rt6i_flags & ~RTF_EXPIRES;
1665 rt->rt6i_metric = 0;
1667 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1668 #ifdef CONFIG_IPV6_SUBTREES
1669 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1671 rt->rt6i_table = ort->rt6i_table;
1676 #ifdef CONFIG_IPV6_ROUTE_INFO
1677 static struct rt6_info *rt6_get_route_info(struct net *net,
1678 struct in6_addr *prefix, int prefixlen,
1679 struct in6_addr *gwaddr, int ifindex)
1681 struct fib6_node *fn;
1682 struct rt6_info *rt = NULL;
1683 struct fib6_table *table;
1685 table = fib6_get_table(net, RT6_TABLE_INFO);
1689 write_lock_bh(&table->tb6_lock);
1690 fn = fib6_locate(&table->tb6_root, prefix ,prefixlen, NULL, 0);
1694 for (rt = fn->leaf; rt; rt = rt->u.dst.rt6_next) {
1695 if (rt->rt6i_dev->ifindex != ifindex)
1697 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
1699 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
1701 dst_hold(&rt->u.dst);
1705 write_unlock_bh(&table->tb6_lock);
1709 static struct rt6_info *rt6_add_route_info(struct net *net,
1710 struct in6_addr *prefix, int prefixlen,
1711 struct in6_addr *gwaddr, int ifindex,
1714 struct fib6_config cfg = {
1715 .fc_table = RT6_TABLE_INFO,
1716 .fc_metric = IP6_RT_PRIO_USER,
1717 .fc_ifindex = ifindex,
1718 .fc_dst_len = prefixlen,
1719 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
1720 RTF_UP | RTF_PREF(pref),
1722 .fc_nlinfo.nlh = NULL,
1723 .fc_nlinfo.nl_net = net,
1726 ipv6_addr_copy(&cfg.fc_dst, prefix);
1727 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1729 /* We should treat it as a default route if prefix length is 0. */
1731 cfg.fc_flags |= RTF_DEFAULT;
1733 ip6_route_add(&cfg);
1735 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
1739 struct rt6_info *rt6_get_dflt_router(struct in6_addr *addr, struct net_device *dev)
1741 struct rt6_info *rt;
1742 struct fib6_table *table;
1744 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
1748 write_lock_bh(&table->tb6_lock);
1749 for (rt = table->tb6_root.leaf; rt; rt=rt->u.dst.rt6_next) {
1750 if (dev == rt->rt6i_dev &&
1751 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
1752 ipv6_addr_equal(&rt->rt6i_gateway, addr))
1756 dst_hold(&rt->u.dst);
1757 write_unlock_bh(&table->tb6_lock);
1761 struct rt6_info *rt6_add_dflt_router(struct in6_addr *gwaddr,
1762 struct net_device *dev,
1765 struct fib6_config cfg = {
1766 .fc_table = RT6_TABLE_DFLT,
1767 .fc_metric = IP6_RT_PRIO_USER,
1768 .fc_ifindex = dev->ifindex,
1769 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
1770 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
1772 .fc_nlinfo.nlh = NULL,
1773 .fc_nlinfo.nl_net = dev_net(dev),
1776 ipv6_addr_copy(&cfg.fc_gateway, gwaddr);
1778 ip6_route_add(&cfg);
1780 return rt6_get_dflt_router(gwaddr, dev);
1783 void rt6_purge_dflt_routers(struct net *net)
1785 struct rt6_info *rt;
1786 struct fib6_table *table;
1788 /* NOTE: Keep consistent with rt6_get_dflt_router */
1789 table = fib6_get_table(net, RT6_TABLE_DFLT);
1794 read_lock_bh(&table->tb6_lock);
1795 for (rt = table->tb6_root.leaf; rt; rt = rt->u.dst.rt6_next) {
1796 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF)) {
1797 dst_hold(&rt->u.dst);
1798 read_unlock_bh(&table->tb6_lock);
1803 read_unlock_bh(&table->tb6_lock);
1806 static void rtmsg_to_fib6_config(struct net *net,
1807 struct in6_rtmsg *rtmsg,
1808 struct fib6_config *cfg)
1810 memset(cfg, 0, sizeof(*cfg));
1812 cfg->fc_table = RT6_TABLE_MAIN;
1813 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
1814 cfg->fc_metric = rtmsg->rtmsg_metric;
1815 cfg->fc_expires = rtmsg->rtmsg_info;
1816 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
1817 cfg->fc_src_len = rtmsg->rtmsg_src_len;
1818 cfg->fc_flags = rtmsg->rtmsg_flags;
1820 cfg->fc_nlinfo.nl_net = net;
1822 ipv6_addr_copy(&cfg->fc_dst, &rtmsg->rtmsg_dst);
1823 ipv6_addr_copy(&cfg->fc_src, &rtmsg->rtmsg_src);
1824 ipv6_addr_copy(&cfg->fc_gateway, &rtmsg->rtmsg_gateway);
1827 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
1829 struct fib6_config cfg;
1830 struct in6_rtmsg rtmsg;
1834 case SIOCADDRT: /* Add a route */
1835 case SIOCDELRT: /* Delete a route */
1836 if (!capable(CAP_NET_ADMIN))
1838 err = copy_from_user(&rtmsg, arg,
1839 sizeof(struct in6_rtmsg));
1843 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
1848 err = ip6_route_add(&cfg);
1851 err = ip6_route_del(&cfg);
1865 * Drop the packet on the floor
1868 static int ip6_pkt_drop(struct sk_buff *skb, int code, int ipstats_mib_noroutes)
1871 struct dst_entry *dst = skb_dst(skb);
1872 switch (ipstats_mib_noroutes) {
1873 case IPSTATS_MIB_INNOROUTES:
1874 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
1875 if (type == IPV6_ADDR_ANY || type == IPV6_ADDR_RESERVED) {
1876 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1877 IPSTATS_MIB_INADDRERRORS);
1881 case IPSTATS_MIB_OUTNOROUTES:
1882 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
1883 ipstats_mib_noroutes);
1886 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0, skb->dev);
1891 static int ip6_pkt_discard(struct sk_buff *skb)
1893 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
1896 static int ip6_pkt_discard_out(struct sk_buff *skb)
1898 skb->dev = skb_dst(skb)->dev;
1899 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
1902 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1904 static int ip6_pkt_prohibit(struct sk_buff *skb)
1906 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
1909 static int ip6_pkt_prohibit_out(struct sk_buff *skb)
1911 skb->dev = skb_dst(skb)->dev;
1912 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
1918 * Allocate a dst for local (unicast / anycast) address.
1921 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
1922 const struct in6_addr *addr,
1925 struct net *net = dev_net(idev->dev);
1926 struct rt6_info *rt = ip6_dst_alloc(net->ipv6.ip6_dst_ops);
1927 struct neighbour *neigh;
1930 return ERR_PTR(-ENOMEM);
1932 dev_hold(net->loopback_dev);
1935 rt->u.dst.flags = DST_HOST;
1936 rt->u.dst.input = ip6_input;
1937 rt->u.dst.output = ip6_output;
1938 rt->rt6i_dev = net->loopback_dev;
1939 rt->rt6i_idev = idev;
1940 rt->u.dst.metrics[RTAX_MTU-1] = ipv6_get_mtu(rt->rt6i_dev);
1941 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, dst_mtu(&rt->u.dst));
1942 rt->u.dst.metrics[RTAX_HOPLIMIT-1] = -1;
1943 rt->u.dst.obsolete = -1;
1945 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
1947 rt->rt6i_flags |= RTF_ANYCAST;
1949 rt->rt6i_flags |= RTF_LOCAL;
1950 neigh = ndisc_get_neigh(rt->rt6i_dev, &rt->rt6i_gateway);
1951 if (IS_ERR(neigh)) {
1952 dst_free(&rt->u.dst);
1954 /* We are casting this because that is the return
1955 * value type. But an errno encoded pointer is the
1956 * same regardless of the underlying pointer type,
1957 * and that's what we are returning. So this is OK.
1959 return (struct rt6_info *) neigh;
1961 rt->rt6i_nexthop = neigh;
1963 ipv6_addr_copy(&rt->rt6i_dst.addr, addr);
1964 rt->rt6i_dst.plen = 128;
1965 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
1967 atomic_set(&rt->u.dst.__refcnt, 1);
1972 struct arg_dev_net {
1973 struct net_device *dev;
1977 static int fib6_ifdown(struct rt6_info *rt, void *arg)
1979 struct net_device *dev = ((struct arg_dev_net *)arg)->dev;
1980 struct net *net = ((struct arg_dev_net *)arg)->net;
1982 if (((void *)rt->rt6i_dev == dev || dev == NULL) &&
1983 rt != net->ipv6.ip6_null_entry) {
1984 RT6_TRACE("deleted by ifdown %p\n", rt);
1990 void rt6_ifdown(struct net *net, struct net_device *dev)
1992 struct arg_dev_net adn = {
1997 fib6_clean_all(net, fib6_ifdown, 0, &adn);
1998 icmp6_clean_all(fib6_ifdown, &adn);
2001 struct rt6_mtu_change_arg
2003 struct net_device *dev;
2007 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2009 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2010 struct inet6_dev *idev;
2011 struct net *net = dev_net(arg->dev);
2013 /* In IPv6 pmtu discovery is not optional,
2014 so that RTAX_MTU lock cannot disable it.
2015 We still use this lock to block changes
2016 caused by addrconf/ndisc.
2019 idev = __in6_dev_get(arg->dev);
2023 /* For administrative MTU increase, there is no way to discover
2024 IPv6 PMTU increase, so PMTU increase should be updated here.
2025 Since RFC 1981 doesn't include administrative MTU increase
2026 update PMTU increase is a MUST. (i.e. jumbo frame)
2029 If new MTU is less than route PMTU, this new MTU will be the
2030 lowest MTU in the path, update the route PMTU to reflect PMTU
2031 decreases; if new MTU is greater than route PMTU, and the
2032 old MTU is the lowest MTU in the path, update the route PMTU
2033 to reflect the increase. In this case if the other nodes' MTU
2034 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2037 if (rt->rt6i_dev == arg->dev &&
2038 !dst_metric_locked(&rt->u.dst, RTAX_MTU) &&
2039 (dst_mtu(&rt->u.dst) >= arg->mtu ||
2040 (dst_mtu(&rt->u.dst) < arg->mtu &&
2041 dst_mtu(&rt->u.dst) == idev->cnf.mtu6))) {
2042 rt->u.dst.metrics[RTAX_MTU-1] = arg->mtu;
2043 rt->u.dst.metrics[RTAX_ADVMSS-1] = ipv6_advmss(net, arg->mtu);
2048 void rt6_mtu_change(struct net_device *dev, unsigned mtu)
2050 struct rt6_mtu_change_arg arg = {
2055 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, 0, &arg);
2058 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2059 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2060 [RTA_OIF] = { .type = NLA_U32 },
2061 [RTA_IIF] = { .type = NLA_U32 },
2062 [RTA_PRIORITY] = { .type = NLA_U32 },
2063 [RTA_METRICS] = { .type = NLA_NESTED },
2066 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2067 struct fib6_config *cfg)
2070 struct nlattr *tb[RTA_MAX+1];
2073 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2078 rtm = nlmsg_data(nlh);
2079 memset(cfg, 0, sizeof(*cfg));
2081 cfg->fc_table = rtm->rtm_table;
2082 cfg->fc_dst_len = rtm->rtm_dst_len;
2083 cfg->fc_src_len = rtm->rtm_src_len;
2084 cfg->fc_flags = RTF_UP;
2085 cfg->fc_protocol = rtm->rtm_protocol;
2087 if (rtm->rtm_type == RTN_UNREACHABLE)
2088 cfg->fc_flags |= RTF_REJECT;
2090 cfg->fc_nlinfo.pid = NETLINK_CB(skb).pid;
2091 cfg->fc_nlinfo.nlh = nlh;
2092 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2094 if (tb[RTA_GATEWAY]) {
2095 nla_memcpy(&cfg->fc_gateway, tb[RTA_GATEWAY], 16);
2096 cfg->fc_flags |= RTF_GATEWAY;
2100 int plen = (rtm->rtm_dst_len + 7) >> 3;
2102 if (nla_len(tb[RTA_DST]) < plen)
2105 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2109 int plen = (rtm->rtm_src_len + 7) >> 3;
2111 if (nla_len(tb[RTA_SRC]) < plen)
2114 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2118 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2120 if (tb[RTA_PRIORITY])
2121 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2123 if (tb[RTA_METRICS]) {
2124 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2125 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2129 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2136 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2138 struct fib6_config cfg;
2141 err = rtm_to_fib6_config(skb, nlh, &cfg);
2145 return ip6_route_del(&cfg);
2148 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr* nlh, void *arg)
2150 struct fib6_config cfg;
2153 err = rtm_to_fib6_config(skb, nlh, &cfg);
2157 return ip6_route_add(&cfg);
2160 static inline size_t rt6_nlmsg_size(void)
2162 return NLMSG_ALIGN(sizeof(struct rtmsg))
2163 + nla_total_size(16) /* RTA_SRC */
2164 + nla_total_size(16) /* RTA_DST */
2165 + nla_total_size(16) /* RTA_GATEWAY */
2166 + nla_total_size(16) /* RTA_PREFSRC */
2167 + nla_total_size(4) /* RTA_TABLE */
2168 + nla_total_size(4) /* RTA_IIF */
2169 + nla_total_size(4) /* RTA_OIF */
2170 + nla_total_size(4) /* RTA_PRIORITY */
2171 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
2172 + nla_total_size(sizeof(struct rta_cacheinfo));
2175 static int rt6_fill_node(struct net *net,
2176 struct sk_buff *skb, struct rt6_info *rt,
2177 struct in6_addr *dst, struct in6_addr *src,
2178 int iif, int type, u32 pid, u32 seq,
2179 int prefix, int nowait, unsigned int flags)
2182 struct nlmsghdr *nlh;
2186 if (prefix) { /* user wants prefix routes only */
2187 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
2188 /* success since this is not a prefix route */
2193 nlh = nlmsg_put(skb, pid, seq, type, sizeof(*rtm), flags);
2197 rtm = nlmsg_data(nlh);
2198 rtm->rtm_family = AF_INET6;
2199 rtm->rtm_dst_len = rt->rt6i_dst.plen;
2200 rtm->rtm_src_len = rt->rt6i_src.plen;
2203 table = rt->rt6i_table->tb6_id;
2205 table = RT6_TABLE_UNSPEC;
2206 rtm->rtm_table = table;
2207 NLA_PUT_U32(skb, RTA_TABLE, table);
2208 if (rt->rt6i_flags&RTF_REJECT)
2209 rtm->rtm_type = RTN_UNREACHABLE;
2210 else if (rt->rt6i_dev && (rt->rt6i_dev->flags&IFF_LOOPBACK))
2211 rtm->rtm_type = RTN_LOCAL;
2213 rtm->rtm_type = RTN_UNICAST;
2215 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
2216 rtm->rtm_protocol = rt->rt6i_protocol;
2217 if (rt->rt6i_flags&RTF_DYNAMIC)
2218 rtm->rtm_protocol = RTPROT_REDIRECT;
2219 else if (rt->rt6i_flags & RTF_ADDRCONF)
2220 rtm->rtm_protocol = RTPROT_KERNEL;
2221 else if (rt->rt6i_flags&RTF_DEFAULT)
2222 rtm->rtm_protocol = RTPROT_RA;
2224 if (rt->rt6i_flags&RTF_CACHE)
2225 rtm->rtm_flags |= RTM_F_CLONED;
2228 NLA_PUT(skb, RTA_DST, 16, dst);
2229 rtm->rtm_dst_len = 128;
2230 } else if (rtm->rtm_dst_len)
2231 NLA_PUT(skb, RTA_DST, 16, &rt->rt6i_dst.addr);
2232 #ifdef CONFIG_IPV6_SUBTREES
2234 NLA_PUT(skb, RTA_SRC, 16, src);
2235 rtm->rtm_src_len = 128;
2236 } else if (rtm->rtm_src_len)
2237 NLA_PUT(skb, RTA_SRC, 16, &rt->rt6i_src.addr);
2240 #ifdef CONFIG_IPV6_MROUTE
2241 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
2242 int err = ip6mr_get_route(net, skb, rtm, nowait);
2247 goto nla_put_failure;
2249 if (err == -EMSGSIZE)
2250 goto nla_put_failure;
2255 NLA_PUT_U32(skb, RTA_IIF, iif);
2257 struct inet6_dev *idev = ip6_dst_idev(&rt->u.dst);
2258 struct in6_addr saddr_buf;
2259 if (ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2260 dst, 0, &saddr_buf) == 0)
2261 NLA_PUT(skb, RTA_PREFSRC, 16, &saddr_buf);
2264 if (rtnetlink_put_metrics(skb, rt->u.dst.metrics) < 0)
2265 goto nla_put_failure;
2267 if (rt->u.dst.neighbour)
2268 NLA_PUT(skb, RTA_GATEWAY, 16, &rt->u.dst.neighbour->primary_key);
2271 NLA_PUT_U32(skb, RTA_OIF, rt->rt6i_dev->ifindex);
2273 NLA_PUT_U32(skb, RTA_PRIORITY, rt->rt6i_metric);
2275 if (!(rt->rt6i_flags & RTF_EXPIRES))
2277 else if (rt->rt6i_expires - jiffies < INT_MAX)
2278 expires = rt->rt6i_expires - jiffies;
2282 if (rtnl_put_cacheinfo(skb, &rt->u.dst, 0, 0, 0,
2283 expires, rt->u.dst.error) < 0)
2284 goto nla_put_failure;
2286 return nlmsg_end(skb, nlh);
2289 nlmsg_cancel(skb, nlh);
2293 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
2295 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
2298 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
2299 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
2300 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
2304 return rt6_fill_node(arg->net,
2305 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
2306 NETLINK_CB(arg->cb->skb).pid, arg->cb->nlh->nlmsg_seq,
2307 prefix, 0, NLM_F_MULTI);
2310 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr* nlh, void *arg)
2312 struct net *net = sock_net(in_skb->sk);
2313 struct nlattr *tb[RTA_MAX+1];
2314 struct rt6_info *rt;
2315 struct sk_buff *skb;
2320 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2325 memset(&fl, 0, sizeof(fl));
2328 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
2331 ipv6_addr_copy(&fl.fl6_src, nla_data(tb[RTA_SRC]));
2335 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
2338 ipv6_addr_copy(&fl.fl6_dst, nla_data(tb[RTA_DST]));
2342 iif = nla_get_u32(tb[RTA_IIF]);
2345 fl.oif = nla_get_u32(tb[RTA_OIF]);
2348 struct net_device *dev;
2349 dev = __dev_get_by_index(net, iif);
2356 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
2362 /* Reserve room for dummy headers, this skb can pass
2363 through good chunk of routing engine.
2365 skb_reset_mac_header(skb);
2366 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
2368 rt = (struct rt6_info*) ip6_route_output(net, NULL, &fl);
2369 skb_dst_set(skb, &rt->u.dst);
2371 err = rt6_fill_node(net, skb, rt, &fl.fl6_dst, &fl.fl6_src, iif,
2372 RTM_NEWROUTE, NETLINK_CB(in_skb).pid,
2373 nlh->nlmsg_seq, 0, 0, 0);
2379 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
2384 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info)
2386 struct sk_buff *skb;
2387 struct net *net = info->nl_net;
2392 seq = info->nlh != NULL ? info->nlh->nlmsg_seq : 0;
2394 skb = nlmsg_new(rt6_nlmsg_size(), gfp_any());
2398 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
2399 event, info->pid, seq, 0, 0, 0);
2401 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2402 WARN_ON(err == -EMSGSIZE);
2406 rtnl_notify(skb, net, info->pid, RTNLGRP_IPV6_ROUTE,
2407 info->nlh, gfp_any());
2411 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
2414 static int ip6_route_dev_notify(struct notifier_block *this,
2415 unsigned long event, void *data)
2417 struct net_device *dev = (struct net_device *)data;
2418 struct net *net = dev_net(dev);
2420 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
2421 net->ipv6.ip6_null_entry->u.dst.dev = dev;
2422 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
2423 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2424 net->ipv6.ip6_prohibit_entry->u.dst.dev = dev;
2425 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
2426 net->ipv6.ip6_blk_hole_entry->u.dst.dev = dev;
2427 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
2438 #ifdef CONFIG_PROC_FS
2440 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2451 static int rt6_info_route(struct rt6_info *rt, void *p_arg)
2453 struct seq_file *m = p_arg;
2455 seq_printf(m, "%pi6 %02x ", &rt->rt6i_dst.addr, rt->rt6i_dst.plen);
2457 #ifdef CONFIG_IPV6_SUBTREES
2458 seq_printf(m, "%pi6 %02x ", &rt->rt6i_src.addr, rt->rt6i_src.plen);
2460 seq_puts(m, "00000000000000000000000000000000 00 ");
2463 if (rt->rt6i_nexthop) {
2464 seq_printf(m, "%pi6", rt->rt6i_nexthop->primary_key);
2466 seq_puts(m, "00000000000000000000000000000000");
2468 seq_printf(m, " %08x %08x %08x %08x %8s\n",
2469 rt->rt6i_metric, atomic_read(&rt->u.dst.__refcnt),
2470 rt->u.dst.__use, rt->rt6i_flags,
2471 rt->rt6i_dev ? rt->rt6i_dev->name : "");
2475 static int ipv6_route_show(struct seq_file *m, void *v)
2477 struct net *net = (struct net *)m->private;
2478 fib6_clean_all(net, rt6_info_route, 0, m);
2482 static int ipv6_route_open(struct inode *inode, struct file *file)
2484 return single_open_net(inode, file, ipv6_route_show);
2487 static const struct file_operations ipv6_route_proc_fops = {
2488 .owner = THIS_MODULE,
2489 .open = ipv6_route_open,
2491 .llseek = seq_lseek,
2492 .release = single_release_net,
2495 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
2497 struct net *net = (struct net *)seq->private;
2498 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
2499 net->ipv6.rt6_stats->fib_nodes,
2500 net->ipv6.rt6_stats->fib_route_nodes,
2501 net->ipv6.rt6_stats->fib_rt_alloc,
2502 net->ipv6.rt6_stats->fib_rt_entries,
2503 net->ipv6.rt6_stats->fib_rt_cache,
2504 atomic_read(&net->ipv6.ip6_dst_ops->entries),
2505 net->ipv6.rt6_stats->fib_discarded_routes);
2510 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
2512 return single_open_net(inode, file, rt6_stats_seq_show);
2515 static const struct file_operations rt6_stats_seq_fops = {
2516 .owner = THIS_MODULE,
2517 .open = rt6_stats_seq_open,
2519 .llseek = seq_lseek,
2520 .release = single_release_net,
2522 #endif /* CONFIG_PROC_FS */
2524 #ifdef CONFIG_SYSCTL
2527 int ipv6_sysctl_rtcache_flush(ctl_table *ctl, int write, struct file * filp,
2528 void __user *buffer, size_t *lenp, loff_t *ppos)
2530 struct net *net = current->nsproxy->net_ns;
2531 int delay = net->ipv6.sysctl.flush_delay;
2533 proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
2534 fib6_run_gc(delay <= 0 ? ~0UL : (unsigned long)delay, net);
2540 ctl_table ipv6_route_table_template[] = {
2542 .procname = "flush",
2543 .data = &init_net.ipv6.sysctl.flush_delay,
2544 .maxlen = sizeof(int),
2546 .proc_handler = ipv6_sysctl_rtcache_flush
2549 .ctl_name = NET_IPV6_ROUTE_GC_THRESH,
2550 .procname = "gc_thresh",
2551 .data = &ip6_dst_ops_template.gc_thresh,
2552 .maxlen = sizeof(int),
2554 .proc_handler = proc_dointvec,
2557 .ctl_name = NET_IPV6_ROUTE_MAX_SIZE,
2558 .procname = "max_size",
2559 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
2560 .maxlen = sizeof(int),
2562 .proc_handler = proc_dointvec,
2565 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL,
2566 .procname = "gc_min_interval",
2567 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2568 .maxlen = sizeof(int),
2570 .proc_handler = proc_dointvec_jiffies,
2571 .strategy = sysctl_jiffies,
2574 .ctl_name = NET_IPV6_ROUTE_GC_TIMEOUT,
2575 .procname = "gc_timeout",
2576 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
2577 .maxlen = sizeof(int),
2579 .proc_handler = proc_dointvec_jiffies,
2580 .strategy = sysctl_jiffies,
2583 .ctl_name = NET_IPV6_ROUTE_GC_INTERVAL,
2584 .procname = "gc_interval",
2585 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
2586 .maxlen = sizeof(int),
2588 .proc_handler = proc_dointvec_jiffies,
2589 .strategy = sysctl_jiffies,
2592 .ctl_name = NET_IPV6_ROUTE_GC_ELASTICITY,
2593 .procname = "gc_elasticity",
2594 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
2595 .maxlen = sizeof(int),
2597 .proc_handler = proc_dointvec_jiffies,
2598 .strategy = sysctl_jiffies,
2601 .ctl_name = NET_IPV6_ROUTE_MTU_EXPIRES,
2602 .procname = "mtu_expires",
2603 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
2604 .maxlen = sizeof(int),
2606 .proc_handler = proc_dointvec_jiffies,
2607 .strategy = sysctl_jiffies,
2610 .ctl_name = NET_IPV6_ROUTE_MIN_ADVMSS,
2611 .procname = "min_adv_mss",
2612 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
2613 .maxlen = sizeof(int),
2615 .proc_handler = proc_dointvec_jiffies,
2616 .strategy = sysctl_jiffies,
2619 .ctl_name = NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS,
2620 .procname = "gc_min_interval_ms",
2621 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
2622 .maxlen = sizeof(int),
2624 .proc_handler = proc_dointvec_ms_jiffies,
2625 .strategy = sysctl_ms_jiffies,
2630 struct ctl_table *ipv6_route_sysctl_init(struct net *net)
2632 struct ctl_table *table;
2634 table = kmemdup(ipv6_route_table_template,
2635 sizeof(ipv6_route_table_template),
2639 table[0].data = &net->ipv6.sysctl.flush_delay;
2640 table[1].data = &net->ipv6.ip6_dst_ops->gc_thresh;
2641 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
2642 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
2643 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
2644 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
2645 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
2646 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
2647 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
2654 static int ip6_route_net_init(struct net *net)
2658 net->ipv6.ip6_dst_ops = kmemdup(&ip6_dst_ops_template,
2659 sizeof(*net->ipv6.ip6_dst_ops),
2661 if (!net->ipv6.ip6_dst_ops)
2663 net->ipv6.ip6_dst_ops->dst_net = hold_net(net);
2665 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
2666 sizeof(*net->ipv6.ip6_null_entry),
2668 if (!net->ipv6.ip6_null_entry)
2669 goto out_ip6_dst_ops;
2670 net->ipv6.ip6_null_entry->u.dst.path =
2671 (struct dst_entry *)net->ipv6.ip6_null_entry;
2672 net->ipv6.ip6_null_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2674 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2675 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
2676 sizeof(*net->ipv6.ip6_prohibit_entry),
2678 if (!net->ipv6.ip6_prohibit_entry)
2679 goto out_ip6_null_entry;
2680 net->ipv6.ip6_prohibit_entry->u.dst.path =
2681 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
2682 net->ipv6.ip6_prohibit_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2684 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
2685 sizeof(*net->ipv6.ip6_blk_hole_entry),
2687 if (!net->ipv6.ip6_blk_hole_entry)
2688 goto out_ip6_prohibit_entry;
2689 net->ipv6.ip6_blk_hole_entry->u.dst.path =
2690 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
2691 net->ipv6.ip6_blk_hole_entry->u.dst.ops = net->ipv6.ip6_dst_ops;
2694 net->ipv6.sysctl.flush_delay = 0;
2695 net->ipv6.sysctl.ip6_rt_max_size = 4096;
2696 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
2697 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
2698 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
2699 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
2700 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
2701 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
2703 #ifdef CONFIG_PROC_FS
2704 proc_net_fops_create(net, "ipv6_route", 0, &ipv6_route_proc_fops);
2705 proc_net_fops_create(net, "rt6_stats", S_IRUGO, &rt6_stats_seq_fops);
2707 net->ipv6.ip6_rt_gc_expire = 30*HZ;
2713 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2714 out_ip6_prohibit_entry:
2715 kfree(net->ipv6.ip6_prohibit_entry);
2717 kfree(net->ipv6.ip6_null_entry);
2720 release_net(net->ipv6.ip6_dst_ops->dst_net);
2721 kfree(net->ipv6.ip6_dst_ops);
2725 static void ip6_route_net_exit(struct net *net)
2727 #ifdef CONFIG_PROC_FS
2728 proc_net_remove(net, "ipv6_route");
2729 proc_net_remove(net, "rt6_stats");
2731 kfree(net->ipv6.ip6_null_entry);
2732 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2733 kfree(net->ipv6.ip6_prohibit_entry);
2734 kfree(net->ipv6.ip6_blk_hole_entry);
2736 release_net(net->ipv6.ip6_dst_ops->dst_net);
2737 kfree(net->ipv6.ip6_dst_ops);
2740 static struct pernet_operations ip6_route_net_ops = {
2741 .init = ip6_route_net_init,
2742 .exit = ip6_route_net_exit,
2745 static struct notifier_block ip6_route_dev_notifier = {
2746 .notifier_call = ip6_route_dev_notify,
2750 int __init ip6_route_init(void)
2755 ip6_dst_ops_template.kmem_cachep =
2756 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
2757 SLAB_HWCACHE_ALIGN, NULL);
2758 if (!ip6_dst_ops_template.kmem_cachep)
2761 ret = register_pernet_subsys(&ip6_route_net_ops);
2763 goto out_kmem_cache;
2765 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
2767 /* Registering of the loopback is done before this portion of code,
2768 * the loopback reference in rt6_info will not be taken, do it
2769 * manually for init_net */
2770 init_net.ipv6.ip6_null_entry->u.dst.dev = init_net.loopback_dev;
2771 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2772 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2773 init_net.ipv6.ip6_prohibit_entry->u.dst.dev = init_net.loopback_dev;
2774 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2775 init_net.ipv6.ip6_blk_hole_entry->u.dst.dev = init_net.loopback_dev;
2776 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
2780 goto out_register_subsys;
2786 ret = fib6_rules_init();
2791 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL) ||
2792 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL) ||
2793 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL))
2794 goto fib6_rules_init;
2796 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
2798 goto fib6_rules_init;
2804 fib6_rules_cleanup();
2809 out_register_subsys:
2810 unregister_pernet_subsys(&ip6_route_net_ops);
2812 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
2816 void ip6_route_cleanup(void)
2818 unregister_netdevice_notifier(&ip6_route_dev_notifier);
2819 fib6_rules_cleanup();
2822 unregister_pernet_subsys(&ip6_route_net_ops);
2823 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);