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 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
57 #include <net/dst_metadata.h>
59 #include <net/netevent.h>
60 #include <net/netlink.h>
61 #include <net/nexthop.h>
62 #include <net/lwtunnel.h>
63 #include <net/ip_tunnels.h>
64 #include <net/l3mdev.h>
66 #include <asm/uaccess.h>
69 #include <linux/sysctl.h>
73 RT6_NUD_FAIL_HARD = -3,
74 RT6_NUD_FAIL_PROBE = -2,
75 RT6_NUD_FAIL_DO_RR = -1,
79 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
80 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
81 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
82 static unsigned int ip6_mtu(const struct dst_entry *dst);
83 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
84 static void ip6_dst_destroy(struct dst_entry *);
85 static void ip6_dst_ifdown(struct dst_entry *,
86 struct net_device *dev, int how);
87 static int ip6_dst_gc(struct dst_ops *ops);
89 static int ip6_pkt_discard(struct sk_buff *skb);
90 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
91 static int ip6_pkt_prohibit(struct sk_buff *skb);
92 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
93 static void ip6_link_failure(struct sk_buff *skb);
94 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
95 struct sk_buff *skb, u32 mtu);
96 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
98 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
99 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
101 #ifdef CONFIG_IPV6_ROUTE_INFO
102 static struct rt6_info *rt6_add_route_info(struct net *net,
103 const struct in6_addr *prefix, int prefixlen,
104 const struct in6_addr *gwaddr, int ifindex,
106 static struct rt6_info *rt6_get_route_info(struct net *net,
107 const struct in6_addr *prefix, int prefixlen,
108 const struct in6_addr *gwaddr, int ifindex);
111 struct uncached_list {
113 struct list_head head;
116 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
118 static void rt6_uncached_list_add(struct rt6_info *rt)
120 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
122 rt->dst.flags |= DST_NOCACHE;
123 rt->rt6i_uncached_list = ul;
125 spin_lock_bh(&ul->lock);
126 list_add_tail(&rt->rt6i_uncached, &ul->head);
127 spin_unlock_bh(&ul->lock);
130 static void rt6_uncached_list_del(struct rt6_info *rt)
132 if (!list_empty(&rt->rt6i_uncached)) {
133 struct uncached_list *ul = rt->rt6i_uncached_list;
135 spin_lock_bh(&ul->lock);
136 list_del(&rt->rt6i_uncached);
137 spin_unlock_bh(&ul->lock);
141 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
143 struct net_device *loopback_dev = net->loopback_dev;
146 if (dev == loopback_dev)
149 for_each_possible_cpu(cpu) {
150 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
153 spin_lock_bh(&ul->lock);
154 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
155 struct inet6_dev *rt_idev = rt->rt6i_idev;
156 struct net_device *rt_dev = rt->dst.dev;
158 if (rt_idev->dev == dev) {
159 rt->rt6i_idev = in6_dev_get(loopback_dev);
160 in6_dev_put(rt_idev);
164 rt->dst.dev = loopback_dev;
165 dev_hold(rt->dst.dev);
169 spin_unlock_bh(&ul->lock);
173 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
175 return dst_metrics_write_ptr(rt->dst.from);
178 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
180 struct rt6_info *rt = (struct rt6_info *)dst;
182 if (rt->rt6i_flags & RTF_PCPU)
183 return rt6_pcpu_cow_metrics(rt);
184 else if (rt->rt6i_flags & RTF_CACHE)
187 return dst_cow_metrics_generic(dst, old);
190 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
194 struct in6_addr *p = &rt->rt6i_gateway;
196 if (!ipv6_addr_any(p))
197 return (const void *) p;
199 return &ipv6_hdr(skb)->daddr;
203 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
207 struct rt6_info *rt = (struct rt6_info *) dst;
210 daddr = choose_neigh_daddr(rt, skb, daddr);
211 n = __ipv6_neigh_lookup(dst->dev, daddr);
214 return neigh_create(&nd_tbl, daddr, dst->dev);
217 static struct dst_ops ip6_dst_ops_template = {
221 .check = ip6_dst_check,
222 .default_advmss = ip6_default_advmss,
224 .cow_metrics = ipv6_cow_metrics,
225 .destroy = ip6_dst_destroy,
226 .ifdown = ip6_dst_ifdown,
227 .negative_advice = ip6_negative_advice,
228 .link_failure = ip6_link_failure,
229 .update_pmtu = ip6_rt_update_pmtu,
230 .redirect = rt6_do_redirect,
231 .local_out = __ip6_local_out,
232 .neigh_lookup = ip6_neigh_lookup,
235 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
237 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
239 return mtu ? : dst->dev->mtu;
242 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
243 struct sk_buff *skb, u32 mtu)
247 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
252 static struct dst_ops ip6_dst_blackhole_ops = {
254 .destroy = ip6_dst_destroy,
255 .check = ip6_dst_check,
256 .mtu = ip6_blackhole_mtu,
257 .default_advmss = ip6_default_advmss,
258 .update_pmtu = ip6_rt_blackhole_update_pmtu,
259 .redirect = ip6_rt_blackhole_redirect,
260 .cow_metrics = dst_cow_metrics_generic,
261 .neigh_lookup = ip6_neigh_lookup,
264 static const u32 ip6_template_metrics[RTAX_MAX] = {
265 [RTAX_HOPLIMIT - 1] = 0,
268 static const struct rt6_info ip6_null_entry_template = {
270 .__refcnt = ATOMIC_INIT(1),
272 .obsolete = DST_OBSOLETE_FORCE_CHK,
273 .error = -ENETUNREACH,
274 .input = ip6_pkt_discard,
275 .output = ip6_pkt_discard_out,
277 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
278 .rt6i_protocol = RTPROT_KERNEL,
279 .rt6i_metric = ~(u32) 0,
280 .rt6i_ref = ATOMIC_INIT(1),
283 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
285 static const struct rt6_info ip6_prohibit_entry_template = {
287 .__refcnt = ATOMIC_INIT(1),
289 .obsolete = DST_OBSOLETE_FORCE_CHK,
291 .input = ip6_pkt_prohibit,
292 .output = ip6_pkt_prohibit_out,
294 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
295 .rt6i_protocol = RTPROT_KERNEL,
296 .rt6i_metric = ~(u32) 0,
297 .rt6i_ref = ATOMIC_INIT(1),
300 static const struct rt6_info ip6_blk_hole_entry_template = {
302 .__refcnt = ATOMIC_INIT(1),
304 .obsolete = DST_OBSOLETE_FORCE_CHK,
306 .input = dst_discard,
307 .output = dst_discard_out,
309 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
310 .rt6i_protocol = RTPROT_KERNEL,
311 .rt6i_metric = ~(u32) 0,
312 .rt6i_ref = ATOMIC_INIT(1),
317 static void rt6_info_init(struct rt6_info *rt)
319 struct dst_entry *dst = &rt->dst;
321 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
322 INIT_LIST_HEAD(&rt->rt6i_siblings);
323 INIT_LIST_HEAD(&rt->rt6i_uncached);
326 /* allocate dst with ip6_dst_ops */
327 static struct rt6_info *__ip6_dst_alloc(struct net *net,
328 struct net_device *dev,
331 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
332 0, DST_OBSOLETE_FORCE_CHK, flags);
340 static struct rt6_info *ip6_dst_alloc(struct net *net,
341 struct net_device *dev,
344 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
347 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
351 for_each_possible_cpu(cpu) {
354 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
355 /* no one shares rt */
359 dst_destroy((struct dst_entry *)rt);
367 static void ip6_dst_destroy(struct dst_entry *dst)
369 struct rt6_info *rt = (struct rt6_info *)dst;
370 struct dst_entry *from = dst->from;
371 struct inet6_dev *idev;
373 dst_destroy_metrics_generic(dst);
374 free_percpu(rt->rt6i_pcpu);
375 rt6_uncached_list_del(rt);
377 idev = rt->rt6i_idev;
379 rt->rt6i_idev = NULL;
387 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
390 struct rt6_info *rt = (struct rt6_info *)dst;
391 struct inet6_dev *idev = rt->rt6i_idev;
392 struct net_device *loopback_dev =
393 dev_net(dev)->loopback_dev;
395 if (dev != loopback_dev) {
396 if (idev && idev->dev == dev) {
397 struct inet6_dev *loopback_idev =
398 in6_dev_get(loopback_dev);
400 rt->rt6i_idev = loopback_idev;
407 static bool rt6_check_expired(const struct rt6_info *rt)
409 if (rt->rt6i_flags & RTF_EXPIRES) {
410 if (time_after(jiffies, rt->dst.expires))
412 } else if (rt->dst.from) {
413 return rt6_check_expired((struct rt6_info *) rt->dst.from);
418 /* Multipath route selection:
419 * Hash based function using packet header and flowlabel.
420 * Adapted from fib_info_hashfn()
422 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
423 const struct flowi6 *fl6)
425 return get_hash_from_flowi6(fl6) % candidate_count;
428 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
429 struct flowi6 *fl6, int oif,
432 struct rt6_info *sibling, *next_sibling;
435 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
436 /* Don't change the route, if route_choosen == 0
437 * (siblings does not include ourself)
440 list_for_each_entry_safe(sibling, next_sibling,
441 &match->rt6i_siblings, rt6i_siblings) {
443 if (route_choosen == 0) {
444 if (rt6_score_route(sibling, oif, strict) < 0)
454 * Route lookup. Any table->tb6_lock is implied.
457 static inline struct rt6_info *rt6_device_match(struct net *net,
459 const struct in6_addr *saddr,
463 struct rt6_info *local = NULL;
464 struct rt6_info *sprt;
466 if (!oif && ipv6_addr_any(saddr))
469 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
470 struct net_device *dev = sprt->dst.dev;
473 if (dev->ifindex == oif)
475 if (dev->flags & IFF_LOOPBACK) {
476 if (!sprt->rt6i_idev ||
477 sprt->rt6i_idev->dev->ifindex != oif) {
478 if (flags & RT6_LOOKUP_F_IFACE)
481 local->rt6i_idev->dev->ifindex == oif)
487 if (ipv6_chk_addr(net, saddr, dev,
488 flags & RT6_LOOKUP_F_IFACE))
497 if (flags & RT6_LOOKUP_F_IFACE)
498 return net->ipv6.ip6_null_entry;
504 #ifdef CONFIG_IPV6_ROUTER_PREF
505 struct __rt6_probe_work {
506 struct work_struct work;
507 struct in6_addr target;
508 struct net_device *dev;
511 static void rt6_probe_deferred(struct work_struct *w)
513 struct in6_addr mcaddr;
514 struct __rt6_probe_work *work =
515 container_of(w, struct __rt6_probe_work, work);
517 addrconf_addr_solict_mult(&work->target, &mcaddr);
518 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, NULL);
523 static void rt6_probe(struct rt6_info *rt)
525 struct __rt6_probe_work *work;
526 struct neighbour *neigh;
528 * Okay, this does not seem to be appropriate
529 * for now, however, we need to check if it
530 * is really so; aka Router Reachability Probing.
532 * Router Reachability Probe MUST be rate-limited
533 * to no more than one per minute.
535 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
538 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
540 if (neigh->nud_state & NUD_VALID)
544 write_lock(&neigh->lock);
545 if (!(neigh->nud_state & NUD_VALID) &&
548 rt->rt6i_idev->cnf.rtr_probe_interval)) {
549 work = kmalloc(sizeof(*work), GFP_ATOMIC);
551 __neigh_set_probe_once(neigh);
553 write_unlock(&neigh->lock);
555 work = kmalloc(sizeof(*work), GFP_ATOMIC);
559 INIT_WORK(&work->work, rt6_probe_deferred);
560 work->target = rt->rt6i_gateway;
561 dev_hold(rt->dst.dev);
562 work->dev = rt->dst.dev;
563 schedule_work(&work->work);
567 rcu_read_unlock_bh();
570 static inline void rt6_probe(struct rt6_info *rt)
576 * Default Router Selection (RFC 2461 6.3.6)
578 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
580 struct net_device *dev = rt->dst.dev;
581 if (!oif || dev->ifindex == oif)
583 if ((dev->flags & IFF_LOOPBACK) &&
584 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
589 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
591 struct neighbour *neigh;
592 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
594 if (rt->rt6i_flags & RTF_NONEXTHOP ||
595 !(rt->rt6i_flags & RTF_GATEWAY))
596 return RT6_NUD_SUCCEED;
599 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
601 read_lock(&neigh->lock);
602 if (neigh->nud_state & NUD_VALID)
603 ret = RT6_NUD_SUCCEED;
604 #ifdef CONFIG_IPV6_ROUTER_PREF
605 else if (!(neigh->nud_state & NUD_FAILED))
606 ret = RT6_NUD_SUCCEED;
608 ret = RT6_NUD_FAIL_PROBE;
610 read_unlock(&neigh->lock);
612 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
613 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
615 rcu_read_unlock_bh();
620 static int rt6_score_route(struct rt6_info *rt, int oif,
625 m = rt6_check_dev(rt, oif);
626 if (!m && (strict & RT6_LOOKUP_F_IFACE))
627 return RT6_NUD_FAIL_HARD;
628 #ifdef CONFIG_IPV6_ROUTER_PREF
629 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
631 if (strict & RT6_LOOKUP_F_REACHABLE) {
632 int n = rt6_check_neigh(rt);
639 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
640 int *mpri, struct rt6_info *match,
644 bool match_do_rr = false;
645 struct inet6_dev *idev = rt->rt6i_idev;
646 struct net_device *dev = rt->dst.dev;
648 if (dev && !netif_carrier_ok(dev) &&
649 idev->cnf.ignore_routes_with_linkdown)
652 if (rt6_check_expired(rt))
655 m = rt6_score_route(rt, oif, strict);
656 if (m == RT6_NUD_FAIL_DO_RR) {
658 m = 0; /* lowest valid score */
659 } else if (m == RT6_NUD_FAIL_HARD) {
663 if (strict & RT6_LOOKUP_F_REACHABLE)
666 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
668 *do_rr = match_do_rr;
676 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
677 struct rt6_info *rr_head,
678 u32 metric, int oif, int strict,
681 struct rt6_info *rt, *match, *cont;
686 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
687 if (rt->rt6i_metric != metric) {
692 match = find_match(rt, oif, strict, &mpri, match, do_rr);
695 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
696 if (rt->rt6i_metric != metric) {
701 match = find_match(rt, oif, strict, &mpri, match, do_rr);
707 for (rt = cont; rt; rt = rt->dst.rt6_next)
708 match = find_match(rt, oif, strict, &mpri, match, do_rr);
713 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
715 struct rt6_info *match, *rt0;
721 fn->rr_ptr = rt0 = fn->leaf;
723 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
727 struct rt6_info *next = rt0->dst.rt6_next;
729 /* no entries matched; do round-robin */
730 if (!next || next->rt6i_metric != rt0->rt6i_metric)
737 net = dev_net(rt0->dst.dev);
738 return match ? match : net->ipv6.ip6_null_entry;
741 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
743 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
746 #ifdef CONFIG_IPV6_ROUTE_INFO
747 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
748 const struct in6_addr *gwaddr)
750 struct net *net = dev_net(dev);
751 struct route_info *rinfo = (struct route_info *) opt;
752 struct in6_addr prefix_buf, *prefix;
754 unsigned long lifetime;
757 if (len < sizeof(struct route_info)) {
761 /* Sanity check for prefix_len and length */
762 if (rinfo->length > 3) {
764 } else if (rinfo->prefix_len > 128) {
766 } else if (rinfo->prefix_len > 64) {
767 if (rinfo->length < 2) {
770 } else if (rinfo->prefix_len > 0) {
771 if (rinfo->length < 1) {
776 pref = rinfo->route_pref;
777 if (pref == ICMPV6_ROUTER_PREF_INVALID)
780 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
782 if (rinfo->length == 3)
783 prefix = (struct in6_addr *)rinfo->prefix;
785 /* this function is safe */
786 ipv6_addr_prefix(&prefix_buf,
787 (struct in6_addr *)rinfo->prefix,
789 prefix = &prefix_buf;
792 if (rinfo->prefix_len == 0)
793 rt = rt6_get_dflt_router(gwaddr, dev);
795 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
796 gwaddr, dev->ifindex);
798 if (rt && !lifetime) {
804 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
807 rt->rt6i_flags = RTF_ROUTEINFO |
808 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
811 if (!addrconf_finite_timeout(lifetime))
812 rt6_clean_expires(rt);
814 rt6_set_expires(rt, jiffies + HZ * lifetime);
822 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
823 struct in6_addr *saddr)
825 struct fib6_node *pn;
827 if (fn->fn_flags & RTN_TL_ROOT)
830 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
831 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
834 if (fn->fn_flags & RTN_RTINFO)
839 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
840 struct fib6_table *table,
841 struct flowi6 *fl6, int flags)
843 struct fib6_node *fn;
846 read_lock_bh(&table->tb6_lock);
847 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
850 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
851 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
852 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
853 if (rt == net->ipv6.ip6_null_entry) {
854 fn = fib6_backtrack(fn, &fl6->saddr);
858 dst_use(&rt->dst, jiffies);
859 read_unlock_bh(&table->tb6_lock);
864 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
867 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
869 EXPORT_SYMBOL_GPL(ip6_route_lookup);
871 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
872 const struct in6_addr *saddr, int oif, int strict)
874 struct flowi6 fl6 = {
878 struct dst_entry *dst;
879 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
882 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
883 flags |= RT6_LOOKUP_F_HAS_SADDR;
886 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
888 return (struct rt6_info *) dst;
894 EXPORT_SYMBOL(rt6_lookup);
896 /* ip6_ins_rt is called with FREE table->tb6_lock.
897 It takes new route entry, the addition fails by any reason the
898 route is freed. In any case, if caller does not hold it, it may
902 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
903 struct mx6_config *mxc)
906 struct fib6_table *table;
908 table = rt->rt6i_table;
909 write_lock_bh(&table->tb6_lock);
910 err = fib6_add(&table->tb6_root, rt, info, mxc);
911 write_unlock_bh(&table->tb6_lock);
916 int ip6_ins_rt(struct rt6_info *rt)
918 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
919 struct mx6_config mxc = { .mx = NULL, };
921 return __ip6_ins_rt(rt, &info, &mxc);
924 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
925 const struct in6_addr *daddr,
926 const struct in6_addr *saddr)
934 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
935 ort = (struct rt6_info *)ort->dst.from;
937 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
942 ip6_rt_copy_init(rt, ort);
943 rt->rt6i_flags |= RTF_CACHE;
945 rt->dst.flags |= DST_HOST;
946 rt->rt6i_dst.addr = *daddr;
947 rt->rt6i_dst.plen = 128;
949 if (!rt6_is_gw_or_nonexthop(ort)) {
950 if (ort->rt6i_dst.plen != 128 &&
951 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
952 rt->rt6i_flags |= RTF_ANYCAST;
953 #ifdef CONFIG_IPV6_SUBTREES
954 if (rt->rt6i_src.plen && saddr) {
955 rt->rt6i_src.addr = *saddr;
956 rt->rt6i_src.plen = 128;
964 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
966 struct rt6_info *pcpu_rt;
968 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
969 rt->dst.dev, rt->dst.flags);
973 ip6_rt_copy_init(pcpu_rt, rt);
974 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
975 pcpu_rt->rt6i_flags |= RTF_PCPU;
979 /* It should be called with read_lock_bh(&tb6_lock) acquired */
980 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
982 struct rt6_info *pcpu_rt, **p;
984 p = this_cpu_ptr(rt->rt6i_pcpu);
988 dst_hold(&pcpu_rt->dst);
989 rt6_dst_from_metrics_check(pcpu_rt);
994 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
996 struct fib6_table *table = rt->rt6i_table;
997 struct rt6_info *pcpu_rt, *prev, **p;
999 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1001 struct net *net = dev_net(rt->dst.dev);
1003 dst_hold(&net->ipv6.ip6_null_entry->dst);
1004 return net->ipv6.ip6_null_entry;
1007 read_lock_bh(&table->tb6_lock);
1008 if (rt->rt6i_pcpu) {
1009 p = this_cpu_ptr(rt->rt6i_pcpu);
1010 prev = cmpxchg(p, NULL, pcpu_rt);
1012 /* If someone did it before us, return prev instead */
1013 dst_destroy(&pcpu_rt->dst);
1017 /* rt has been removed from the fib6 tree
1018 * before we have a chance to acquire the read_lock.
1019 * In this case, don't brother to create a pcpu rt
1020 * since rt is going away anyway. The next
1021 * dst_check() will trigger a re-lookup.
1023 dst_destroy(&pcpu_rt->dst);
1026 dst_hold(&pcpu_rt->dst);
1027 rt6_dst_from_metrics_check(pcpu_rt);
1028 read_unlock_bh(&table->tb6_lock);
1032 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1033 struct flowi6 *fl6, int flags)
1035 struct fib6_node *fn, *saved_fn;
1036 struct rt6_info *rt;
1039 strict |= flags & RT6_LOOKUP_F_IFACE;
1040 if (net->ipv6.devconf_all->forwarding == 0)
1041 strict |= RT6_LOOKUP_F_REACHABLE;
1043 read_lock_bh(&table->tb6_lock);
1045 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1048 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1052 rt = rt6_select(fn, oif, strict);
1053 if (rt->rt6i_nsiblings)
1054 rt = rt6_multipath_select(rt, fl6, oif, strict);
1055 if (rt == net->ipv6.ip6_null_entry) {
1056 fn = fib6_backtrack(fn, &fl6->saddr);
1058 goto redo_rt6_select;
1059 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1060 /* also consider unreachable route */
1061 strict &= ~RT6_LOOKUP_F_REACHABLE;
1063 goto redo_rt6_select;
1068 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1069 dst_use(&rt->dst, jiffies);
1070 read_unlock_bh(&table->tb6_lock);
1072 rt6_dst_from_metrics_check(rt);
1074 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1075 !(rt->rt6i_flags & RTF_GATEWAY))) {
1076 /* Create a RTF_CACHE clone which will not be
1077 * owned by the fib6 tree. It is for the special case where
1078 * the daddr in the skb during the neighbor look-up is different
1079 * from the fl6->daddr used to look-up route here.
1082 struct rt6_info *uncached_rt;
1084 dst_use(&rt->dst, jiffies);
1085 read_unlock_bh(&table->tb6_lock);
1087 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1088 dst_release(&rt->dst);
1091 rt6_uncached_list_add(uncached_rt);
1093 uncached_rt = net->ipv6.ip6_null_entry;
1095 dst_hold(&uncached_rt->dst);
1099 /* Get a percpu copy */
1101 struct rt6_info *pcpu_rt;
1103 rt->dst.lastuse = jiffies;
1105 pcpu_rt = rt6_get_pcpu_route(rt);
1108 read_unlock_bh(&table->tb6_lock);
1110 /* We have to do the read_unlock first
1111 * because rt6_make_pcpu_route() may trigger
1112 * ip6_dst_gc() which will take the write_lock.
1115 read_unlock_bh(&table->tb6_lock);
1116 pcpu_rt = rt6_make_pcpu_route(rt);
1117 dst_release(&rt->dst);
1125 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1126 struct flowi6 *fl6, int flags)
1128 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1131 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1132 struct net_device *dev,
1133 struct flowi6 *fl6, int flags)
1135 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1136 flags |= RT6_LOOKUP_F_IFACE;
1138 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1141 void ip6_route_input(struct sk_buff *skb)
1143 const struct ipv6hdr *iph = ipv6_hdr(skb);
1144 struct net *net = dev_net(skb->dev);
1145 int flags = RT6_LOOKUP_F_HAS_SADDR;
1146 struct ip_tunnel_info *tun_info;
1147 struct flowi6 fl6 = {
1148 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1149 .daddr = iph->daddr,
1150 .saddr = iph->saddr,
1151 .flowlabel = ip6_flowinfo(iph),
1152 .flowi6_mark = skb->mark,
1153 .flowi6_proto = iph->nexthdr,
1156 tun_info = skb_tunnel_info(skb);
1157 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1158 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1160 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1163 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1164 struct flowi6 *fl6, int flags)
1166 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1169 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1172 struct dst_entry *dst;
1176 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1180 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1182 any_src = ipv6_addr_any(&fl6->saddr);
1183 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1184 (fl6->flowi6_oif && any_src))
1185 flags |= RT6_LOOKUP_F_IFACE;
1188 flags |= RT6_LOOKUP_F_HAS_SADDR;
1190 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1192 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1194 EXPORT_SYMBOL(ip6_route_output);
1196 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1198 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1199 struct dst_entry *new = NULL;
1201 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1207 new->input = dst_discard;
1208 new->output = dst_discard_out;
1210 dst_copy_metrics(new, &ort->dst);
1211 rt->rt6i_idev = ort->rt6i_idev;
1213 in6_dev_hold(rt->rt6i_idev);
1215 rt->rt6i_gateway = ort->rt6i_gateway;
1216 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1217 rt->rt6i_metric = 0;
1219 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1220 #ifdef CONFIG_IPV6_SUBTREES
1221 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1227 dst_release(dst_orig);
1228 return new ? new : ERR_PTR(-ENOMEM);
1232 * Destination cache support functions
1235 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1238 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1239 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1242 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1244 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1247 if (rt6_check_expired(rt))
1253 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1255 if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1256 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1262 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1264 struct rt6_info *rt;
1266 rt = (struct rt6_info *) dst;
1268 /* All IPV6 dsts are created with ->obsolete set to the value
1269 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1270 * into this function always.
1273 rt6_dst_from_metrics_check(rt);
1275 if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1276 return rt6_dst_from_check(rt, cookie);
1278 return rt6_check(rt, cookie);
1281 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1283 struct rt6_info *rt = (struct rt6_info *) dst;
1286 if (rt->rt6i_flags & RTF_CACHE) {
1287 if (rt6_check_expired(rt)) {
1299 static void ip6_link_failure(struct sk_buff *skb)
1301 struct rt6_info *rt;
1303 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1305 rt = (struct rt6_info *) skb_dst(skb);
1307 if (rt->rt6i_flags & RTF_CACHE) {
1310 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1311 rt->rt6i_node->fn_sernum = -1;
1316 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1318 struct net *net = dev_net(rt->dst.dev);
1320 rt->rt6i_flags |= RTF_MODIFIED;
1321 rt->rt6i_pmtu = mtu;
1322 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1325 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1326 const struct ipv6hdr *iph, u32 mtu)
1328 struct rt6_info *rt6 = (struct rt6_info *)dst;
1330 if (rt6->rt6i_flags & RTF_LOCAL)
1334 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1335 if (mtu >= dst_mtu(dst))
1338 if (rt6->rt6i_flags & RTF_CACHE) {
1339 rt6_do_update_pmtu(rt6, mtu);
1341 const struct in6_addr *daddr, *saddr;
1342 struct rt6_info *nrt6;
1345 daddr = &iph->daddr;
1346 saddr = &iph->saddr;
1348 daddr = &sk->sk_v6_daddr;
1349 saddr = &inet6_sk(sk)->saddr;
1353 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1355 rt6_do_update_pmtu(nrt6, mtu);
1357 /* ip6_ins_rt(nrt6) will bump the
1358 * rt6->rt6i_node->fn_sernum
1359 * which will fail the next rt6_check() and
1360 * invalidate the sk->sk_dst_cache.
1367 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1368 struct sk_buff *skb, u32 mtu)
1370 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1373 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1376 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1377 struct dst_entry *dst;
1380 memset(&fl6, 0, sizeof(fl6));
1381 fl6.flowi6_oif = oif;
1382 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1383 fl6.daddr = iph->daddr;
1384 fl6.saddr = iph->saddr;
1385 fl6.flowlabel = ip6_flowinfo(iph);
1387 dst = ip6_route_output(net, NULL, &fl6);
1389 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1392 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1394 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1396 ip6_update_pmtu(skb, sock_net(sk), mtu,
1397 sk->sk_bound_dev_if, sk->sk_mark);
1399 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1401 /* Handle redirects */
1402 struct ip6rd_flowi {
1404 struct in6_addr gateway;
1407 static struct rt6_info *__ip6_route_redirect(struct net *net,
1408 struct fib6_table *table,
1412 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1413 struct rt6_info *rt;
1414 struct fib6_node *fn;
1416 /* Get the "current" route for this destination and
1417 * check if the redirect has come from approriate router.
1419 * RFC 4861 specifies that redirects should only be
1420 * accepted if they come from the nexthop to the target.
1421 * Due to the way the routes are chosen, this notion
1422 * is a bit fuzzy and one might need to check all possible
1426 read_lock_bh(&table->tb6_lock);
1427 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1429 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1430 if (rt6_check_expired(rt))
1434 if (!(rt->rt6i_flags & RTF_GATEWAY))
1436 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1438 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1444 rt = net->ipv6.ip6_null_entry;
1445 else if (rt->dst.error) {
1446 rt = net->ipv6.ip6_null_entry;
1450 if (rt == net->ipv6.ip6_null_entry) {
1451 fn = fib6_backtrack(fn, &fl6->saddr);
1459 read_unlock_bh(&table->tb6_lock);
1464 static struct dst_entry *ip6_route_redirect(struct net *net,
1465 const struct flowi6 *fl6,
1466 const struct in6_addr *gateway)
1468 int flags = RT6_LOOKUP_F_HAS_SADDR;
1469 struct ip6rd_flowi rdfl;
1472 rdfl.gateway = *gateway;
1474 return fib6_rule_lookup(net, &rdfl.fl6,
1475 flags, __ip6_route_redirect);
1478 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1480 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1481 struct dst_entry *dst;
1484 memset(&fl6, 0, sizeof(fl6));
1485 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1486 fl6.flowi6_oif = oif;
1487 fl6.flowi6_mark = mark;
1488 fl6.daddr = iph->daddr;
1489 fl6.saddr = iph->saddr;
1490 fl6.flowlabel = ip6_flowinfo(iph);
1492 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1493 rt6_do_redirect(dst, NULL, skb);
1496 EXPORT_SYMBOL_GPL(ip6_redirect);
1498 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1501 const struct ipv6hdr *iph = ipv6_hdr(skb);
1502 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1503 struct dst_entry *dst;
1506 memset(&fl6, 0, sizeof(fl6));
1507 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1508 fl6.flowi6_oif = oif;
1509 fl6.flowi6_mark = mark;
1510 fl6.daddr = msg->dest;
1511 fl6.saddr = iph->daddr;
1513 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1514 rt6_do_redirect(dst, NULL, skb);
1518 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1520 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1522 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1524 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1526 struct net_device *dev = dst->dev;
1527 unsigned int mtu = dst_mtu(dst);
1528 struct net *net = dev_net(dev);
1530 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1532 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1533 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1536 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1537 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1538 * IPV6_MAXPLEN is also valid and means: "any MSS,
1539 * rely only on pmtu discovery"
1541 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1546 static unsigned int ip6_mtu(const struct dst_entry *dst)
1548 const struct rt6_info *rt = (const struct rt6_info *)dst;
1549 unsigned int mtu = rt->rt6i_pmtu;
1550 struct inet6_dev *idev;
1555 mtu = dst_metric_raw(dst, RTAX_MTU);
1562 idev = __in6_dev_get(dst->dev);
1564 mtu = idev->cnf.mtu6;
1568 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1571 static struct dst_entry *icmp6_dst_gc_list;
1572 static DEFINE_SPINLOCK(icmp6_dst_lock);
1574 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1577 struct dst_entry *dst;
1578 struct rt6_info *rt;
1579 struct inet6_dev *idev = in6_dev_get(dev);
1580 struct net *net = dev_net(dev);
1582 if (unlikely(!idev))
1583 return ERR_PTR(-ENODEV);
1585 rt = ip6_dst_alloc(net, dev, 0);
1586 if (unlikely(!rt)) {
1588 dst = ERR_PTR(-ENOMEM);
1592 rt->dst.flags |= DST_HOST;
1593 rt->dst.output = ip6_output;
1594 atomic_set(&rt->dst.__refcnt, 1);
1595 rt->rt6i_gateway = fl6->daddr;
1596 rt->rt6i_dst.addr = fl6->daddr;
1597 rt->rt6i_dst.plen = 128;
1598 rt->rt6i_idev = idev;
1599 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1601 spin_lock_bh(&icmp6_dst_lock);
1602 rt->dst.next = icmp6_dst_gc_list;
1603 icmp6_dst_gc_list = &rt->dst;
1604 spin_unlock_bh(&icmp6_dst_lock);
1606 fib6_force_start_gc(net);
1608 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1614 int icmp6_dst_gc(void)
1616 struct dst_entry *dst, **pprev;
1619 spin_lock_bh(&icmp6_dst_lock);
1620 pprev = &icmp6_dst_gc_list;
1622 while ((dst = *pprev) != NULL) {
1623 if (!atomic_read(&dst->__refcnt)) {
1632 spin_unlock_bh(&icmp6_dst_lock);
1637 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1640 struct dst_entry *dst, **pprev;
1642 spin_lock_bh(&icmp6_dst_lock);
1643 pprev = &icmp6_dst_gc_list;
1644 while ((dst = *pprev) != NULL) {
1645 struct rt6_info *rt = (struct rt6_info *) dst;
1646 if (func(rt, arg)) {
1653 spin_unlock_bh(&icmp6_dst_lock);
1656 static int ip6_dst_gc(struct dst_ops *ops)
1658 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1659 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1660 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1661 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1662 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1663 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1666 entries = dst_entries_get_fast(ops);
1667 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1668 entries <= rt_max_size)
1671 net->ipv6.ip6_rt_gc_expire++;
1672 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1673 entries = dst_entries_get_slow(ops);
1674 if (entries < ops->gc_thresh)
1675 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1677 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1678 return entries > rt_max_size;
1681 static int ip6_convert_metrics(struct mx6_config *mxc,
1682 const struct fib6_config *cfg)
1684 bool ecn_ca = false;
1692 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1696 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1697 int type = nla_type(nla);
1702 if (unlikely(type > RTAX_MAX))
1705 if (type == RTAX_CC_ALGO) {
1706 char tmp[TCP_CA_NAME_MAX];
1708 nla_strlcpy(tmp, nla, sizeof(tmp));
1709 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1710 if (val == TCP_CA_UNSPEC)
1713 val = nla_get_u32(nla);
1715 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1719 __set_bit(type - 1, mxc->mx_valid);
1723 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1724 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1734 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1736 struct net *net = cfg->fc_nlinfo.nl_net;
1737 struct rt6_info *rt = NULL;
1738 struct net_device *dev = NULL;
1739 struct inet6_dev *idev = NULL;
1740 struct fib6_table *table;
1744 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1746 #ifndef CONFIG_IPV6_SUBTREES
1747 if (cfg->fc_src_len)
1750 if (cfg->fc_ifindex) {
1752 dev = dev_get_by_index(net, cfg->fc_ifindex);
1755 idev = in6_dev_get(dev);
1760 if (cfg->fc_metric == 0)
1761 cfg->fc_metric = IP6_RT_PRIO_USER;
1764 if (cfg->fc_nlinfo.nlh &&
1765 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1766 table = fib6_get_table(net, cfg->fc_table);
1768 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1769 table = fib6_new_table(net, cfg->fc_table);
1772 table = fib6_new_table(net, cfg->fc_table);
1778 rt = ip6_dst_alloc(net, NULL,
1779 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1786 if (cfg->fc_flags & RTF_EXPIRES)
1787 rt6_set_expires(rt, jiffies +
1788 clock_t_to_jiffies(cfg->fc_expires));
1790 rt6_clean_expires(rt);
1792 if (cfg->fc_protocol == RTPROT_UNSPEC)
1793 cfg->fc_protocol = RTPROT_BOOT;
1794 rt->rt6i_protocol = cfg->fc_protocol;
1796 addr_type = ipv6_addr_type(&cfg->fc_dst);
1798 if (addr_type & IPV6_ADDR_MULTICAST)
1799 rt->dst.input = ip6_mc_input;
1800 else if (cfg->fc_flags & RTF_LOCAL)
1801 rt->dst.input = ip6_input;
1803 rt->dst.input = ip6_forward;
1805 rt->dst.output = ip6_output;
1807 if (cfg->fc_encap) {
1808 struct lwtunnel_state *lwtstate;
1810 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1811 cfg->fc_encap, AF_INET6, cfg,
1815 rt->dst.lwtstate = lwtstate_get(lwtstate);
1816 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1817 rt->dst.lwtstate->orig_output = rt->dst.output;
1818 rt->dst.output = lwtunnel_output;
1820 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1821 rt->dst.lwtstate->orig_input = rt->dst.input;
1822 rt->dst.input = lwtunnel_input;
1826 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1827 rt->rt6i_dst.plen = cfg->fc_dst_len;
1828 if (rt->rt6i_dst.plen == 128)
1829 rt->dst.flags |= DST_HOST;
1831 #ifdef CONFIG_IPV6_SUBTREES
1832 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1833 rt->rt6i_src.plen = cfg->fc_src_len;
1836 rt->rt6i_metric = cfg->fc_metric;
1838 /* We cannot add true routes via loopback here,
1839 they would result in kernel looping; promote them to reject routes
1841 if ((cfg->fc_flags & RTF_REJECT) ||
1842 (dev && (dev->flags & IFF_LOOPBACK) &&
1843 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1844 !(cfg->fc_flags & RTF_LOCAL))) {
1845 /* hold loopback dev/idev if we haven't done so. */
1846 if (dev != net->loopback_dev) {
1851 dev = net->loopback_dev;
1853 idev = in6_dev_get(dev);
1859 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1860 switch (cfg->fc_type) {
1862 rt->dst.error = -EINVAL;
1863 rt->dst.output = dst_discard_out;
1864 rt->dst.input = dst_discard;
1867 rt->dst.error = -EACCES;
1868 rt->dst.output = ip6_pkt_prohibit_out;
1869 rt->dst.input = ip6_pkt_prohibit;
1872 case RTN_UNREACHABLE:
1874 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1875 : (cfg->fc_type == RTN_UNREACHABLE)
1876 ? -EHOSTUNREACH : -ENETUNREACH;
1877 rt->dst.output = ip6_pkt_discard_out;
1878 rt->dst.input = ip6_pkt_discard;
1884 if (cfg->fc_flags & RTF_GATEWAY) {
1885 const struct in6_addr *gw_addr;
1888 gw_addr = &cfg->fc_gateway;
1889 gwa_type = ipv6_addr_type(gw_addr);
1891 /* if gw_addr is local we will fail to detect this in case
1892 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1893 * will return already-added prefix route via interface that
1894 * prefix route was assigned to, which might be non-loopback.
1897 if (ipv6_chk_addr_and_flags(net, gw_addr,
1898 gwa_type & IPV6_ADDR_LINKLOCAL ?
1902 rt->rt6i_gateway = *gw_addr;
1904 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1905 struct rt6_info *grt;
1907 /* IPv6 strictly inhibits using not link-local
1908 addresses as nexthop address.
1909 Otherwise, router will not able to send redirects.
1910 It is very good, but in some (rare!) circumstances
1911 (SIT, PtP, NBMA NOARP links) it is handy to allow
1912 some exceptions. --ANK
1914 if (!(gwa_type & IPV6_ADDR_UNICAST))
1917 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1919 err = -EHOSTUNREACH;
1923 if (dev != grt->dst.dev) {
1929 idev = grt->rt6i_idev;
1931 in6_dev_hold(grt->rt6i_idev);
1933 if (!(grt->rt6i_flags & RTF_GATEWAY))
1941 if (!dev || (dev->flags & IFF_LOOPBACK))
1949 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1950 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1954 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1955 rt->rt6i_prefsrc.plen = 128;
1957 rt->rt6i_prefsrc.plen = 0;
1959 rt->rt6i_flags = cfg->fc_flags;
1963 rt->rt6i_idev = idev;
1964 rt->rt6i_table = table;
1966 cfg->fc_nlinfo.nl_net = dev_net(dev);
1977 return ERR_PTR(err);
1980 int ip6_route_add(struct fib6_config *cfg)
1982 struct mx6_config mxc = { .mx = NULL, };
1983 struct rt6_info *rt;
1986 rt = ip6_route_info_create(cfg);
1993 err = ip6_convert_metrics(&mxc, cfg);
1997 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2009 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2012 struct fib6_table *table;
2013 struct net *net = dev_net(rt->dst.dev);
2015 if (rt == net->ipv6.ip6_null_entry ||
2016 rt->dst.flags & DST_NOCACHE) {
2021 table = rt->rt6i_table;
2022 write_lock_bh(&table->tb6_lock);
2023 err = fib6_del(rt, info);
2024 write_unlock_bh(&table->tb6_lock);
2031 int ip6_del_rt(struct rt6_info *rt)
2033 struct nl_info info = {
2034 .nl_net = dev_net(rt->dst.dev),
2036 return __ip6_del_rt(rt, &info);
2039 static int ip6_route_del(struct fib6_config *cfg)
2041 struct fib6_table *table;
2042 struct fib6_node *fn;
2043 struct rt6_info *rt;
2046 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2050 read_lock_bh(&table->tb6_lock);
2052 fn = fib6_locate(&table->tb6_root,
2053 &cfg->fc_dst, cfg->fc_dst_len,
2054 &cfg->fc_src, cfg->fc_src_len);
2057 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2058 if ((rt->rt6i_flags & RTF_CACHE) &&
2059 !(cfg->fc_flags & RTF_CACHE))
2061 if (cfg->fc_ifindex &&
2063 rt->dst.dev->ifindex != cfg->fc_ifindex))
2065 if (cfg->fc_flags & RTF_GATEWAY &&
2066 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2068 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2071 read_unlock_bh(&table->tb6_lock);
2073 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2076 read_unlock_bh(&table->tb6_lock);
2081 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2083 struct net *net = dev_net(skb->dev);
2084 struct netevent_redirect netevent;
2085 struct rt6_info *rt, *nrt = NULL;
2086 struct ndisc_options ndopts;
2087 struct inet6_dev *in6_dev;
2088 struct neighbour *neigh;
2090 int optlen, on_link;
2093 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2094 optlen -= sizeof(*msg);
2097 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2101 msg = (struct rd_msg *)icmp6_hdr(skb);
2103 if (ipv6_addr_is_multicast(&msg->dest)) {
2104 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2109 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2111 } else if (ipv6_addr_type(&msg->target) !=
2112 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2113 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2117 in6_dev = __in6_dev_get(skb->dev);
2120 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2124 * The IP source address of the Redirect MUST be the same as the current
2125 * first-hop router for the specified ICMP Destination Address.
2128 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2129 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2134 if (ndopts.nd_opts_tgt_lladdr) {
2135 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2138 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2143 rt = (struct rt6_info *) dst;
2144 if (rt == net->ipv6.ip6_null_entry) {
2145 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2149 /* Redirect received -> path was valid.
2150 * Look, redirects are sent only in response to data packets,
2151 * so that this nexthop apparently is reachable. --ANK
2153 dst_confirm(&rt->dst);
2155 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2160 * We have finally decided to accept it.
2163 neigh_update(neigh, lladdr, NUD_STALE,
2164 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2165 NEIGH_UPDATE_F_OVERRIDE|
2166 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2167 NEIGH_UPDATE_F_ISROUTER))
2170 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2174 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2176 nrt->rt6i_flags &= ~RTF_GATEWAY;
2178 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2180 if (ip6_ins_rt(nrt))
2183 netevent.old = &rt->dst;
2184 netevent.new = &nrt->dst;
2185 netevent.daddr = &msg->dest;
2186 netevent.neigh = neigh;
2187 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2189 if (rt->rt6i_flags & RTF_CACHE) {
2190 rt = (struct rt6_info *) dst_clone(&rt->dst);
2195 neigh_release(neigh);
2199 * Misc support functions
2202 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2204 BUG_ON(from->dst.from);
2206 rt->rt6i_flags &= ~RTF_EXPIRES;
2207 dst_hold(&from->dst);
2208 rt->dst.from = &from->dst;
2209 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2212 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2214 rt->dst.input = ort->dst.input;
2215 rt->dst.output = ort->dst.output;
2216 rt->rt6i_dst = ort->rt6i_dst;
2217 rt->dst.error = ort->dst.error;
2218 rt->rt6i_idev = ort->rt6i_idev;
2220 in6_dev_hold(rt->rt6i_idev);
2221 rt->dst.lastuse = jiffies;
2222 rt->rt6i_gateway = ort->rt6i_gateway;
2223 rt->rt6i_flags = ort->rt6i_flags;
2224 rt6_set_from(rt, ort);
2225 rt->rt6i_metric = ort->rt6i_metric;
2226 #ifdef CONFIG_IPV6_SUBTREES
2227 rt->rt6i_src = ort->rt6i_src;
2229 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2230 rt->rt6i_table = ort->rt6i_table;
2231 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2234 #ifdef CONFIG_IPV6_ROUTE_INFO
2235 static struct rt6_info *rt6_get_route_info(struct net *net,
2236 const struct in6_addr *prefix, int prefixlen,
2237 const struct in6_addr *gwaddr, int ifindex)
2239 struct fib6_node *fn;
2240 struct rt6_info *rt = NULL;
2241 struct fib6_table *table;
2243 table = fib6_get_table(net, RT6_TABLE_INFO);
2247 read_lock_bh(&table->tb6_lock);
2248 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2252 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2253 if (rt->dst.dev->ifindex != ifindex)
2255 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2257 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2263 read_unlock_bh(&table->tb6_lock);
2267 static struct rt6_info *rt6_add_route_info(struct net *net,
2268 const struct in6_addr *prefix, int prefixlen,
2269 const struct in6_addr *gwaddr, int ifindex,
2272 struct fib6_config cfg = {
2273 .fc_metric = IP6_RT_PRIO_USER,
2274 .fc_ifindex = ifindex,
2275 .fc_dst_len = prefixlen,
2276 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2277 RTF_UP | RTF_PREF(pref),
2278 .fc_nlinfo.portid = 0,
2279 .fc_nlinfo.nlh = NULL,
2280 .fc_nlinfo.nl_net = net,
2283 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2284 cfg.fc_dst = *prefix;
2285 cfg.fc_gateway = *gwaddr;
2287 /* We should treat it as a default route if prefix length is 0. */
2289 cfg.fc_flags |= RTF_DEFAULT;
2291 ip6_route_add(&cfg);
2293 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2297 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2299 struct rt6_info *rt;
2300 struct fib6_table *table;
2302 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2306 read_lock_bh(&table->tb6_lock);
2307 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2308 if (dev == rt->dst.dev &&
2309 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2310 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2315 read_unlock_bh(&table->tb6_lock);
2319 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2320 struct net_device *dev,
2323 struct fib6_config cfg = {
2324 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2325 .fc_metric = IP6_RT_PRIO_USER,
2326 .fc_ifindex = dev->ifindex,
2327 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2328 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2329 .fc_nlinfo.portid = 0,
2330 .fc_nlinfo.nlh = NULL,
2331 .fc_nlinfo.nl_net = dev_net(dev),
2334 cfg.fc_gateway = *gwaddr;
2336 ip6_route_add(&cfg);
2338 return rt6_get_dflt_router(gwaddr, dev);
2341 void rt6_purge_dflt_routers(struct net *net)
2343 struct rt6_info *rt;
2344 struct fib6_table *table;
2346 /* NOTE: Keep consistent with rt6_get_dflt_router */
2347 table = fib6_get_table(net, RT6_TABLE_DFLT);
2352 read_lock_bh(&table->tb6_lock);
2353 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2354 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2355 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2357 read_unlock_bh(&table->tb6_lock);
2362 read_unlock_bh(&table->tb6_lock);
2365 static void rtmsg_to_fib6_config(struct net *net,
2366 struct in6_rtmsg *rtmsg,
2367 struct fib6_config *cfg)
2369 memset(cfg, 0, sizeof(*cfg));
2371 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2373 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2374 cfg->fc_metric = rtmsg->rtmsg_metric;
2375 cfg->fc_expires = rtmsg->rtmsg_info;
2376 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2377 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2378 cfg->fc_flags = rtmsg->rtmsg_flags;
2380 cfg->fc_nlinfo.nl_net = net;
2382 cfg->fc_dst = rtmsg->rtmsg_dst;
2383 cfg->fc_src = rtmsg->rtmsg_src;
2384 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2387 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2389 struct fib6_config cfg;
2390 struct in6_rtmsg rtmsg;
2394 case SIOCADDRT: /* Add a route */
2395 case SIOCDELRT: /* Delete a route */
2396 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2398 err = copy_from_user(&rtmsg, arg,
2399 sizeof(struct in6_rtmsg));
2403 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2408 err = ip6_route_add(&cfg);
2411 err = ip6_route_del(&cfg);
2425 * Drop the packet on the floor
2428 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2431 struct dst_entry *dst = skb_dst(skb);
2432 switch (ipstats_mib_noroutes) {
2433 case IPSTATS_MIB_INNOROUTES:
2434 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2435 if (type == IPV6_ADDR_ANY) {
2436 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2437 IPSTATS_MIB_INADDRERRORS);
2441 case IPSTATS_MIB_OUTNOROUTES:
2442 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2443 ipstats_mib_noroutes);
2446 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2451 static int ip6_pkt_discard(struct sk_buff *skb)
2453 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2456 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2458 skb->dev = skb_dst(skb)->dev;
2459 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2462 static int ip6_pkt_prohibit(struct sk_buff *skb)
2464 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2467 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2469 skb->dev = skb_dst(skb)->dev;
2470 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2474 * Allocate a dst for local (unicast / anycast) address.
2477 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2478 const struct in6_addr *addr,
2482 struct net *net = dev_net(idev->dev);
2483 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2486 return ERR_PTR(-ENOMEM);
2490 rt->dst.flags |= DST_HOST;
2491 rt->dst.input = ip6_input;
2492 rt->dst.output = ip6_output;
2493 rt->rt6i_idev = idev;
2495 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2497 rt->rt6i_flags |= RTF_ANYCAST;
2499 rt->rt6i_flags |= RTF_LOCAL;
2501 rt->rt6i_gateway = *addr;
2502 rt->rt6i_dst.addr = *addr;
2503 rt->rt6i_dst.plen = 128;
2504 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2505 rt->rt6i_table = fib6_get_table(net, tb_id);
2506 rt->dst.flags |= DST_NOCACHE;
2508 atomic_set(&rt->dst.__refcnt, 1);
2513 int ip6_route_get_saddr(struct net *net,
2514 struct rt6_info *rt,
2515 const struct in6_addr *daddr,
2517 struct in6_addr *saddr)
2519 struct inet6_dev *idev =
2520 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2522 if (rt && rt->rt6i_prefsrc.plen)
2523 *saddr = rt->rt6i_prefsrc.addr;
2525 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2526 daddr, prefs, saddr);
2530 /* remove deleted ip from prefsrc entries */
2531 struct arg_dev_net_ip {
2532 struct net_device *dev;
2534 struct in6_addr *addr;
2537 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2539 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2540 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2541 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2543 if (((void *)rt->dst.dev == dev || !dev) &&
2544 rt != net->ipv6.ip6_null_entry &&
2545 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2546 /* remove prefsrc entry */
2547 rt->rt6i_prefsrc.plen = 0;
2552 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2554 struct net *net = dev_net(ifp->idev->dev);
2555 struct arg_dev_net_ip adni = {
2556 .dev = ifp->idev->dev,
2560 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2563 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2564 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2566 /* Remove routers and update dst entries when gateway turn into host. */
2567 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2569 struct in6_addr *gateway = (struct in6_addr *)arg;
2571 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2572 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2573 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2579 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2581 fib6_clean_all(net, fib6_clean_tohost, gateway);
2584 struct arg_dev_net {
2585 struct net_device *dev;
2589 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2591 const struct arg_dev_net *adn = arg;
2592 const struct net_device *dev = adn->dev;
2594 if ((rt->dst.dev == dev || !dev) &&
2595 rt != adn->net->ipv6.ip6_null_entry)
2601 void rt6_ifdown(struct net *net, struct net_device *dev)
2603 struct arg_dev_net adn = {
2608 fib6_clean_all(net, fib6_ifdown, &adn);
2609 icmp6_clean_all(fib6_ifdown, &adn);
2611 rt6_uncached_list_flush_dev(net, dev);
2614 struct rt6_mtu_change_arg {
2615 struct net_device *dev;
2619 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2621 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2622 struct inet6_dev *idev;
2624 /* In IPv6 pmtu discovery is not optional,
2625 so that RTAX_MTU lock cannot disable it.
2626 We still use this lock to block changes
2627 caused by addrconf/ndisc.
2630 idev = __in6_dev_get(arg->dev);
2634 /* For administrative MTU increase, there is no way to discover
2635 IPv6 PMTU increase, so PMTU increase should be updated here.
2636 Since RFC 1981 doesn't include administrative MTU increase
2637 update PMTU increase is a MUST. (i.e. jumbo frame)
2640 If new MTU is less than route PMTU, this new MTU will be the
2641 lowest MTU in the path, update the route PMTU to reflect PMTU
2642 decreases; if new MTU is greater than route PMTU, and the
2643 old MTU is the lowest MTU in the path, update the route PMTU
2644 to reflect the increase. In this case if the other nodes' MTU
2645 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2648 if (rt->dst.dev == arg->dev &&
2649 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2650 if (rt->rt6i_flags & RTF_CACHE) {
2651 /* For RTF_CACHE with rt6i_pmtu == 0
2652 * (i.e. a redirected route),
2653 * the metrics of its rt->dst.from has already
2656 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2657 rt->rt6i_pmtu = arg->mtu;
2658 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2659 (dst_mtu(&rt->dst) < arg->mtu &&
2660 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2661 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2667 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2669 struct rt6_mtu_change_arg arg = {
2674 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2677 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2678 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2679 [RTA_OIF] = { .type = NLA_U32 },
2680 [RTA_IIF] = { .type = NLA_U32 },
2681 [RTA_PRIORITY] = { .type = NLA_U32 },
2682 [RTA_METRICS] = { .type = NLA_NESTED },
2683 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2684 [RTA_PREF] = { .type = NLA_U8 },
2685 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2686 [RTA_ENCAP] = { .type = NLA_NESTED },
2689 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2690 struct fib6_config *cfg)
2693 struct nlattr *tb[RTA_MAX+1];
2697 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2702 rtm = nlmsg_data(nlh);
2703 memset(cfg, 0, sizeof(*cfg));
2705 cfg->fc_table = rtm->rtm_table;
2706 cfg->fc_dst_len = rtm->rtm_dst_len;
2707 cfg->fc_src_len = rtm->rtm_src_len;
2708 cfg->fc_flags = RTF_UP;
2709 cfg->fc_protocol = rtm->rtm_protocol;
2710 cfg->fc_type = rtm->rtm_type;
2712 if (rtm->rtm_type == RTN_UNREACHABLE ||
2713 rtm->rtm_type == RTN_BLACKHOLE ||
2714 rtm->rtm_type == RTN_PROHIBIT ||
2715 rtm->rtm_type == RTN_THROW)
2716 cfg->fc_flags |= RTF_REJECT;
2718 if (rtm->rtm_type == RTN_LOCAL)
2719 cfg->fc_flags |= RTF_LOCAL;
2721 if (rtm->rtm_flags & RTM_F_CLONED)
2722 cfg->fc_flags |= RTF_CACHE;
2724 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2725 cfg->fc_nlinfo.nlh = nlh;
2726 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2728 if (tb[RTA_GATEWAY]) {
2729 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2730 cfg->fc_flags |= RTF_GATEWAY;
2734 int plen = (rtm->rtm_dst_len + 7) >> 3;
2736 if (nla_len(tb[RTA_DST]) < plen)
2739 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2743 int plen = (rtm->rtm_src_len + 7) >> 3;
2745 if (nla_len(tb[RTA_SRC]) < plen)
2748 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2751 if (tb[RTA_PREFSRC])
2752 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2755 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2757 if (tb[RTA_PRIORITY])
2758 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2760 if (tb[RTA_METRICS]) {
2761 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2762 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2766 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2768 if (tb[RTA_MULTIPATH]) {
2769 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2770 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2774 pref = nla_get_u8(tb[RTA_PREF]);
2775 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2776 pref != ICMPV6_ROUTER_PREF_HIGH)
2777 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2778 cfg->fc_flags |= RTF_PREF(pref);
2782 cfg->fc_encap = tb[RTA_ENCAP];
2784 if (tb[RTA_ENCAP_TYPE])
2785 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2793 struct rt6_info *rt6_info;
2794 struct fib6_config r_cfg;
2795 struct mx6_config mxc;
2796 struct list_head next;
2799 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2803 list_for_each_entry(nh, rt6_nh_list, next) {
2804 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2805 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2806 nh->r_cfg.fc_ifindex);
2810 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2811 struct rt6_info *rt, struct fib6_config *r_cfg)
2814 struct rt6_info *rtnh;
2817 list_for_each_entry(nh, rt6_nh_list, next) {
2818 /* check if rt6_info already exists */
2819 rtnh = nh->rt6_info;
2821 if (rtnh->dst.dev == rt->dst.dev &&
2822 rtnh->rt6i_idev == rt->rt6i_idev &&
2823 ipv6_addr_equal(&rtnh->rt6i_gateway,
2828 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2832 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2837 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2838 list_add_tail(&nh->next, rt6_nh_list);
2843 static int ip6_route_multipath_add(struct fib6_config *cfg)
2845 struct fib6_config r_cfg;
2846 struct rtnexthop *rtnh;
2847 struct rt6_info *rt;
2848 struct rt6_nh *err_nh;
2849 struct rt6_nh *nh, *nh_safe;
2854 int replace = (cfg->fc_nlinfo.nlh &&
2855 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2856 LIST_HEAD(rt6_nh_list);
2858 remaining = cfg->fc_mp_len;
2859 rtnh = (struct rtnexthop *)cfg->fc_mp;
2861 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2862 * rt6_info structs per nexthop
2864 while (rtnh_ok(rtnh, remaining)) {
2865 memcpy(&r_cfg, cfg, sizeof(*cfg));
2866 if (rtnh->rtnh_ifindex)
2867 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2869 attrlen = rtnh_attrlen(rtnh);
2871 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2873 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2875 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2876 r_cfg.fc_flags |= RTF_GATEWAY;
2878 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2879 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2881 r_cfg.fc_encap_type = nla_get_u16(nla);
2884 rt = ip6_route_info_create(&r_cfg);
2891 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2897 rtnh = rtnh_next(rtnh, &remaining);
2901 list_for_each_entry(nh, &rt6_nh_list, next) {
2902 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2903 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2904 nh->rt6_info = NULL;
2907 ip6_print_replace_route_err(&rt6_nh_list);
2912 /* Because each route is added like a single route we remove
2913 * these flags after the first nexthop: if there is a collision,
2914 * we have already failed to add the first nexthop:
2915 * fib6_add_rt2node() has rejected it; when replacing, old
2916 * nexthops have been replaced by first new, the rest should
2919 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2927 /* Delete routes that were already added */
2928 list_for_each_entry(nh, &rt6_nh_list, next) {
2931 ip6_route_del(&nh->r_cfg);
2935 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2937 dst_free(&nh->rt6_info->dst);
2939 list_del(&nh->next);
2946 static int ip6_route_multipath_del(struct fib6_config *cfg)
2948 struct fib6_config r_cfg;
2949 struct rtnexthop *rtnh;
2952 int err = 1, last_err = 0;
2954 remaining = cfg->fc_mp_len;
2955 rtnh = (struct rtnexthop *)cfg->fc_mp;
2957 /* Parse a Multipath Entry */
2958 while (rtnh_ok(rtnh, remaining)) {
2959 memcpy(&r_cfg, cfg, sizeof(*cfg));
2960 if (rtnh->rtnh_ifindex)
2961 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2963 attrlen = rtnh_attrlen(rtnh);
2965 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2967 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2969 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2970 r_cfg.fc_flags |= RTF_GATEWAY;
2973 err = ip6_route_del(&r_cfg);
2977 rtnh = rtnh_next(rtnh, &remaining);
2983 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2985 struct fib6_config cfg;
2988 err = rtm_to_fib6_config(skb, nlh, &cfg);
2993 return ip6_route_multipath_del(&cfg);
2995 return ip6_route_del(&cfg);
2998 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3000 struct fib6_config cfg;
3003 err = rtm_to_fib6_config(skb, nlh, &cfg);
3008 return ip6_route_multipath_add(&cfg);
3010 return ip6_route_add(&cfg);
3013 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3015 return NLMSG_ALIGN(sizeof(struct rtmsg))
3016 + nla_total_size(16) /* RTA_SRC */
3017 + nla_total_size(16) /* RTA_DST */
3018 + nla_total_size(16) /* RTA_GATEWAY */
3019 + nla_total_size(16) /* RTA_PREFSRC */
3020 + nla_total_size(4) /* RTA_TABLE */
3021 + nla_total_size(4) /* RTA_IIF */
3022 + nla_total_size(4) /* RTA_OIF */
3023 + nla_total_size(4) /* RTA_PRIORITY */
3024 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3025 + nla_total_size(sizeof(struct rta_cacheinfo))
3026 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3027 + nla_total_size(1) /* RTA_PREF */
3028 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3031 static int rt6_fill_node(struct net *net,
3032 struct sk_buff *skb, struct rt6_info *rt,
3033 struct in6_addr *dst, struct in6_addr *src,
3034 int iif, int type, u32 portid, u32 seq,
3035 int prefix, int nowait, unsigned int flags)
3037 u32 metrics[RTAX_MAX];
3039 struct nlmsghdr *nlh;
3043 if (prefix) { /* user wants prefix routes only */
3044 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3045 /* success since this is not a prefix route */
3050 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3054 rtm = nlmsg_data(nlh);
3055 rtm->rtm_family = AF_INET6;
3056 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3057 rtm->rtm_src_len = rt->rt6i_src.plen;
3060 table = rt->rt6i_table->tb6_id;
3062 table = RT6_TABLE_UNSPEC;
3063 rtm->rtm_table = table;
3064 if (nla_put_u32(skb, RTA_TABLE, table))
3065 goto nla_put_failure;
3066 if (rt->rt6i_flags & RTF_REJECT) {
3067 switch (rt->dst.error) {
3069 rtm->rtm_type = RTN_BLACKHOLE;
3072 rtm->rtm_type = RTN_PROHIBIT;
3075 rtm->rtm_type = RTN_THROW;
3078 rtm->rtm_type = RTN_UNREACHABLE;
3082 else if (rt->rt6i_flags & RTF_LOCAL)
3083 rtm->rtm_type = RTN_LOCAL;
3084 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3085 rtm->rtm_type = RTN_LOCAL;
3087 rtm->rtm_type = RTN_UNICAST;
3089 if (!netif_carrier_ok(rt->dst.dev)) {
3090 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3091 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3092 rtm->rtm_flags |= RTNH_F_DEAD;
3094 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3095 rtm->rtm_protocol = rt->rt6i_protocol;
3096 if (rt->rt6i_flags & RTF_DYNAMIC)
3097 rtm->rtm_protocol = RTPROT_REDIRECT;
3098 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3099 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3100 rtm->rtm_protocol = RTPROT_RA;
3102 rtm->rtm_protocol = RTPROT_KERNEL;
3105 if (rt->rt6i_flags & RTF_CACHE)
3106 rtm->rtm_flags |= RTM_F_CLONED;
3109 if (nla_put_in6_addr(skb, RTA_DST, dst))
3110 goto nla_put_failure;
3111 rtm->rtm_dst_len = 128;
3112 } else if (rtm->rtm_dst_len)
3113 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3114 goto nla_put_failure;
3115 #ifdef CONFIG_IPV6_SUBTREES
3117 if (nla_put_in6_addr(skb, RTA_SRC, src))
3118 goto nla_put_failure;
3119 rtm->rtm_src_len = 128;
3120 } else if (rtm->rtm_src_len &&
3121 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3122 goto nla_put_failure;
3125 #ifdef CONFIG_IPV6_MROUTE
3126 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3127 int err = ip6mr_get_route(net, skb, rtm, nowait);
3132 goto nla_put_failure;
3134 if (err == -EMSGSIZE)
3135 goto nla_put_failure;
3140 if (nla_put_u32(skb, RTA_IIF, iif))
3141 goto nla_put_failure;
3143 struct in6_addr saddr_buf;
3144 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3145 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3146 goto nla_put_failure;
3149 if (rt->rt6i_prefsrc.plen) {
3150 struct in6_addr saddr_buf;
3151 saddr_buf = rt->rt6i_prefsrc.addr;
3152 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3153 goto nla_put_failure;
3156 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3158 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3159 if (rtnetlink_put_metrics(skb, metrics) < 0)
3160 goto nla_put_failure;
3162 if (rt->rt6i_flags & RTF_GATEWAY) {
3163 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3164 goto nla_put_failure;
3168 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3169 goto nla_put_failure;
3170 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3171 goto nla_put_failure;
3173 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3175 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3176 goto nla_put_failure;
3178 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3179 goto nla_put_failure;
3181 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3183 nlmsg_end(skb, nlh);
3187 nlmsg_cancel(skb, nlh);
3191 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3193 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3196 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3197 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3198 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3202 return rt6_fill_node(arg->net,
3203 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3204 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3205 prefix, 0, NLM_F_MULTI);
3208 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3210 struct net *net = sock_net(in_skb->sk);
3211 struct nlattr *tb[RTA_MAX+1];
3212 struct rt6_info *rt;
3213 struct sk_buff *skb;
3216 int err, iif = 0, oif = 0;
3218 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3223 memset(&fl6, 0, sizeof(fl6));
3226 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3229 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3233 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3236 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3240 iif = nla_get_u32(tb[RTA_IIF]);
3243 oif = nla_get_u32(tb[RTA_OIF]);
3246 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3249 struct net_device *dev;
3252 dev = __dev_get_by_index(net, iif);
3258 fl6.flowi6_iif = iif;
3260 if (!ipv6_addr_any(&fl6.saddr))
3261 flags |= RT6_LOOKUP_F_HAS_SADDR;
3263 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3266 fl6.flowi6_oif = oif;
3268 if (netif_index_is_l3_master(net, oif)) {
3269 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3270 FLOWI_FLAG_SKIP_NH_OIF;
3273 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3276 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3283 /* Reserve room for dummy headers, this skb can pass
3284 through good chunk of routing engine.
3286 skb_reset_mac_header(skb);
3287 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3289 skb_dst_set(skb, &rt->dst);
3291 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3292 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3293 nlh->nlmsg_seq, 0, 0, 0);
3299 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3304 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3305 unsigned int nlm_flags)
3307 struct sk_buff *skb;
3308 struct net *net = info->nl_net;
3313 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3315 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3319 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3320 event, info->portid, seq, 0, 0, nlm_flags);
3322 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3323 WARN_ON(err == -EMSGSIZE);
3327 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3328 info->nlh, gfp_any());
3332 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3335 static int ip6_route_dev_notify(struct notifier_block *this,
3336 unsigned long event, void *ptr)
3338 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3339 struct net *net = dev_net(dev);
3341 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3342 net->ipv6.ip6_null_entry->dst.dev = dev;
3343 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3344 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3345 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3346 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3347 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3348 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3359 #ifdef CONFIG_PROC_FS
3361 static const struct file_operations ipv6_route_proc_fops = {
3362 .owner = THIS_MODULE,
3363 .open = ipv6_route_open,
3365 .llseek = seq_lseek,
3366 .release = seq_release_net,
3369 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3371 struct net *net = (struct net *)seq->private;
3372 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3373 net->ipv6.rt6_stats->fib_nodes,
3374 net->ipv6.rt6_stats->fib_route_nodes,
3375 net->ipv6.rt6_stats->fib_rt_alloc,
3376 net->ipv6.rt6_stats->fib_rt_entries,
3377 net->ipv6.rt6_stats->fib_rt_cache,
3378 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3379 net->ipv6.rt6_stats->fib_discarded_routes);
3384 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3386 return single_open_net(inode, file, rt6_stats_seq_show);
3389 static const struct file_operations rt6_stats_seq_fops = {
3390 .owner = THIS_MODULE,
3391 .open = rt6_stats_seq_open,
3393 .llseek = seq_lseek,
3394 .release = single_release_net,
3396 #endif /* CONFIG_PROC_FS */
3398 #ifdef CONFIG_SYSCTL
3401 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3402 void __user *buffer, size_t *lenp, loff_t *ppos)
3409 net = (struct net *)ctl->extra1;
3410 delay = net->ipv6.sysctl.flush_delay;
3411 proc_dointvec(ctl, write, buffer, lenp, ppos);
3412 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3416 struct ctl_table ipv6_route_table_template[] = {
3418 .procname = "flush",
3419 .data = &init_net.ipv6.sysctl.flush_delay,
3420 .maxlen = sizeof(int),
3422 .proc_handler = ipv6_sysctl_rtcache_flush
3425 .procname = "gc_thresh",
3426 .data = &ip6_dst_ops_template.gc_thresh,
3427 .maxlen = sizeof(int),
3429 .proc_handler = proc_dointvec,
3432 .procname = "max_size",
3433 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3434 .maxlen = sizeof(int),
3436 .proc_handler = proc_dointvec,
3439 .procname = "gc_min_interval",
3440 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3441 .maxlen = sizeof(int),
3443 .proc_handler = proc_dointvec_jiffies,
3446 .procname = "gc_timeout",
3447 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3448 .maxlen = sizeof(int),
3450 .proc_handler = proc_dointvec_jiffies,
3453 .procname = "gc_interval",
3454 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3455 .maxlen = sizeof(int),
3457 .proc_handler = proc_dointvec_jiffies,
3460 .procname = "gc_elasticity",
3461 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3462 .maxlen = sizeof(int),
3464 .proc_handler = proc_dointvec,
3467 .procname = "mtu_expires",
3468 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3469 .maxlen = sizeof(int),
3471 .proc_handler = proc_dointvec_jiffies,
3474 .procname = "min_adv_mss",
3475 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3476 .maxlen = sizeof(int),
3478 .proc_handler = proc_dointvec,
3481 .procname = "gc_min_interval_ms",
3482 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3483 .maxlen = sizeof(int),
3485 .proc_handler = proc_dointvec_ms_jiffies,
3490 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3492 struct ctl_table *table;
3494 table = kmemdup(ipv6_route_table_template,
3495 sizeof(ipv6_route_table_template),
3499 table[0].data = &net->ipv6.sysctl.flush_delay;
3500 table[0].extra1 = net;
3501 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3502 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3503 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3504 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3505 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3506 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3507 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3508 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3509 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3511 /* Don't export sysctls to unprivileged users */
3512 if (net->user_ns != &init_user_ns)
3513 table[0].procname = NULL;
3520 static int __net_init ip6_route_net_init(struct net *net)
3524 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3525 sizeof(net->ipv6.ip6_dst_ops));
3527 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3528 goto out_ip6_dst_ops;
3530 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3531 sizeof(*net->ipv6.ip6_null_entry),
3533 if (!net->ipv6.ip6_null_entry)
3534 goto out_ip6_dst_entries;
3535 net->ipv6.ip6_null_entry->dst.path =
3536 (struct dst_entry *)net->ipv6.ip6_null_entry;
3537 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3538 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3539 ip6_template_metrics, true);
3541 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3542 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3543 sizeof(*net->ipv6.ip6_prohibit_entry),
3545 if (!net->ipv6.ip6_prohibit_entry)
3546 goto out_ip6_null_entry;
3547 net->ipv6.ip6_prohibit_entry->dst.path =
3548 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3549 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3550 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3551 ip6_template_metrics, true);
3553 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3554 sizeof(*net->ipv6.ip6_blk_hole_entry),
3556 if (!net->ipv6.ip6_blk_hole_entry)
3557 goto out_ip6_prohibit_entry;
3558 net->ipv6.ip6_blk_hole_entry->dst.path =
3559 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3560 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3561 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3562 ip6_template_metrics, true);
3565 net->ipv6.sysctl.flush_delay = 0;
3566 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3567 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3568 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3569 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3570 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3571 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3572 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3574 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3580 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3581 out_ip6_prohibit_entry:
3582 kfree(net->ipv6.ip6_prohibit_entry);
3584 kfree(net->ipv6.ip6_null_entry);
3586 out_ip6_dst_entries:
3587 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3592 static void __net_exit ip6_route_net_exit(struct net *net)
3594 kfree(net->ipv6.ip6_null_entry);
3595 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3596 kfree(net->ipv6.ip6_prohibit_entry);
3597 kfree(net->ipv6.ip6_blk_hole_entry);
3599 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3602 static int __net_init ip6_route_net_init_late(struct net *net)
3604 #ifdef CONFIG_PROC_FS
3605 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3606 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3611 static void __net_exit ip6_route_net_exit_late(struct net *net)
3613 #ifdef CONFIG_PROC_FS
3614 remove_proc_entry("ipv6_route", net->proc_net);
3615 remove_proc_entry("rt6_stats", net->proc_net);
3619 static struct pernet_operations ip6_route_net_ops = {
3620 .init = ip6_route_net_init,
3621 .exit = ip6_route_net_exit,
3624 static int __net_init ipv6_inetpeer_init(struct net *net)
3626 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3630 inet_peer_base_init(bp);
3631 net->ipv6.peers = bp;
3635 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3637 struct inet_peer_base *bp = net->ipv6.peers;
3639 net->ipv6.peers = NULL;
3640 inetpeer_invalidate_tree(bp);
3644 static struct pernet_operations ipv6_inetpeer_ops = {
3645 .init = ipv6_inetpeer_init,
3646 .exit = ipv6_inetpeer_exit,
3649 static struct pernet_operations ip6_route_net_late_ops = {
3650 .init = ip6_route_net_init_late,
3651 .exit = ip6_route_net_exit_late,
3654 static struct notifier_block ip6_route_dev_notifier = {
3655 .notifier_call = ip6_route_dev_notify,
3659 int __init ip6_route_init(void)
3665 ip6_dst_ops_template.kmem_cachep =
3666 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3667 SLAB_HWCACHE_ALIGN, NULL);
3668 if (!ip6_dst_ops_template.kmem_cachep)
3671 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3673 goto out_kmem_cache;
3675 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3677 goto out_dst_entries;
3679 ret = register_pernet_subsys(&ip6_route_net_ops);
3681 goto out_register_inetpeer;
3683 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3685 /* Registering of the loopback is done before this portion of code,
3686 * the loopback reference in rt6_info will not be taken, do it
3687 * manually for init_net */
3688 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3689 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3690 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3691 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3692 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3693 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3694 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3698 goto out_register_subsys;
3704 ret = fib6_rules_init();
3708 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3710 goto fib6_rules_init;
3713 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3714 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3715 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3716 goto out_register_late_subsys;
3718 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3720 goto out_register_late_subsys;
3722 for_each_possible_cpu(cpu) {
3723 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3725 INIT_LIST_HEAD(&ul->head);
3726 spin_lock_init(&ul->lock);
3732 out_register_late_subsys:
3733 unregister_pernet_subsys(&ip6_route_net_late_ops);
3735 fib6_rules_cleanup();
3740 out_register_subsys:
3741 unregister_pernet_subsys(&ip6_route_net_ops);
3742 out_register_inetpeer:
3743 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3745 dst_entries_destroy(&ip6_dst_blackhole_ops);
3747 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3751 void ip6_route_cleanup(void)
3753 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3754 unregister_pernet_subsys(&ip6_route_net_late_ops);
3755 fib6_rules_cleanup();
3758 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3759 unregister_pernet_subsys(&ip6_route_net_ops);
3760 dst_entries_destroy(&ip6_dst_blackhole_ops);
3761 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);