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 return time_after(jiffies, rt->dst.expires);
415 static bool rt6_check_expired(const struct rt6_info *rt)
417 if (rt->rt6i_flags & RTF_EXPIRES) {
418 if (time_after(jiffies, rt->dst.expires))
420 } else if (rt->dst.from) {
421 return rt6_check_expired((struct rt6_info *) rt->dst.from);
426 /* Multipath route selection:
427 * Hash based function using packet header and flowlabel.
428 * Adapted from fib_info_hashfn()
430 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
431 const struct flowi6 *fl6)
433 return get_hash_from_flowi6(fl6) % candidate_count;
436 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
437 struct flowi6 *fl6, int oif,
440 struct rt6_info *sibling, *next_sibling;
443 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
444 /* Don't change the route, if route_choosen == 0
445 * (siblings does not include ourself)
448 list_for_each_entry_safe(sibling, next_sibling,
449 &match->rt6i_siblings, rt6i_siblings) {
451 if (route_choosen == 0) {
452 if (rt6_score_route(sibling, oif, strict) < 0)
462 * Route lookup. Any table->tb6_lock is implied.
465 static inline struct rt6_info *rt6_device_match(struct net *net,
467 const struct in6_addr *saddr,
471 struct rt6_info *local = NULL;
472 struct rt6_info *sprt;
474 if (!oif && ipv6_addr_any(saddr))
477 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
478 struct net_device *dev = sprt->dst.dev;
481 if (dev->ifindex == oif)
483 if (dev->flags & IFF_LOOPBACK) {
484 if (!sprt->rt6i_idev ||
485 sprt->rt6i_idev->dev->ifindex != oif) {
486 if (flags & RT6_LOOKUP_F_IFACE)
489 local->rt6i_idev->dev->ifindex == oif)
495 if (ipv6_chk_addr(net, saddr, dev,
496 flags & RT6_LOOKUP_F_IFACE))
505 if (flags & RT6_LOOKUP_F_IFACE)
506 return net->ipv6.ip6_null_entry;
512 #ifdef CONFIG_IPV6_ROUTER_PREF
513 struct __rt6_probe_work {
514 struct work_struct work;
515 struct in6_addr target;
516 struct net_device *dev;
519 static void rt6_probe_deferred(struct work_struct *w)
521 struct in6_addr mcaddr;
522 struct __rt6_probe_work *work =
523 container_of(w, struct __rt6_probe_work, work);
525 addrconf_addr_solict_mult(&work->target, &mcaddr);
526 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
531 static void rt6_probe(struct rt6_info *rt)
533 struct __rt6_probe_work *work;
534 struct neighbour *neigh;
536 * Okay, this does not seem to be appropriate
537 * for now, however, we need to check if it
538 * is really so; aka Router Reachability Probing.
540 * Router Reachability Probe MUST be rate-limited
541 * to no more than one per minute.
543 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
546 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
548 if (neigh->nud_state & NUD_VALID)
552 write_lock(&neigh->lock);
553 if (!(neigh->nud_state & NUD_VALID) &&
556 rt->rt6i_idev->cnf.rtr_probe_interval)) {
557 work = kmalloc(sizeof(*work), GFP_ATOMIC);
559 __neigh_set_probe_once(neigh);
561 write_unlock(&neigh->lock);
563 work = kmalloc(sizeof(*work), GFP_ATOMIC);
567 INIT_WORK(&work->work, rt6_probe_deferred);
568 work->target = rt->rt6i_gateway;
569 dev_hold(rt->dst.dev);
570 work->dev = rt->dst.dev;
571 schedule_work(&work->work);
575 rcu_read_unlock_bh();
578 static inline void rt6_probe(struct rt6_info *rt)
584 * Default Router Selection (RFC 2461 6.3.6)
586 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
588 struct net_device *dev = rt->dst.dev;
589 if (!oif || dev->ifindex == oif)
591 if ((dev->flags & IFF_LOOPBACK) &&
592 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
597 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
599 struct neighbour *neigh;
600 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
602 if (rt->rt6i_flags & RTF_NONEXTHOP ||
603 !(rt->rt6i_flags & RTF_GATEWAY))
604 return RT6_NUD_SUCCEED;
607 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
609 read_lock(&neigh->lock);
610 if (neigh->nud_state & NUD_VALID)
611 ret = RT6_NUD_SUCCEED;
612 #ifdef CONFIG_IPV6_ROUTER_PREF
613 else if (!(neigh->nud_state & NUD_FAILED))
614 ret = RT6_NUD_SUCCEED;
616 ret = RT6_NUD_FAIL_PROBE;
618 read_unlock(&neigh->lock);
620 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
621 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
623 rcu_read_unlock_bh();
628 static int rt6_score_route(struct rt6_info *rt, int oif,
633 m = rt6_check_dev(rt, oif);
634 if (!m && (strict & RT6_LOOKUP_F_IFACE))
635 return RT6_NUD_FAIL_HARD;
636 #ifdef CONFIG_IPV6_ROUTER_PREF
637 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
639 if (strict & RT6_LOOKUP_F_REACHABLE) {
640 int n = rt6_check_neigh(rt);
647 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
648 int *mpri, struct rt6_info *match,
652 bool match_do_rr = false;
653 struct inet6_dev *idev = rt->rt6i_idev;
654 struct net_device *dev = rt->dst.dev;
656 if (dev && !netif_carrier_ok(dev) &&
657 idev->cnf.ignore_routes_with_linkdown)
660 if (rt6_check_expired(rt))
663 m = rt6_score_route(rt, oif, strict);
664 if (m == RT6_NUD_FAIL_DO_RR) {
666 m = 0; /* lowest valid score */
667 } else if (m == RT6_NUD_FAIL_HARD) {
671 if (strict & RT6_LOOKUP_F_REACHABLE)
674 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
676 *do_rr = match_do_rr;
684 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
685 struct rt6_info *rr_head,
686 u32 metric, int oif, int strict,
689 struct rt6_info *rt, *match, *cont;
694 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
695 if (rt->rt6i_metric != metric) {
700 match = find_match(rt, oif, strict, &mpri, match, do_rr);
703 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
704 if (rt->rt6i_metric != metric) {
709 match = find_match(rt, oif, strict, &mpri, match, do_rr);
715 for (rt = cont; rt; rt = rt->dst.rt6_next)
716 match = find_match(rt, oif, strict, &mpri, match, do_rr);
721 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
723 struct rt6_info *match, *rt0;
729 fn->rr_ptr = rt0 = fn->leaf;
731 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
735 struct rt6_info *next = rt0->dst.rt6_next;
737 /* no entries matched; do round-robin */
738 if (!next || next->rt6i_metric != rt0->rt6i_metric)
745 net = dev_net(rt0->dst.dev);
746 return match ? match : net->ipv6.ip6_null_entry;
749 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
751 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
754 #ifdef CONFIG_IPV6_ROUTE_INFO
755 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
756 const struct in6_addr *gwaddr)
758 struct net *net = dev_net(dev);
759 struct route_info *rinfo = (struct route_info *) opt;
760 struct in6_addr prefix_buf, *prefix;
762 unsigned long lifetime;
765 if (len < sizeof(struct route_info)) {
769 /* Sanity check for prefix_len and length */
770 if (rinfo->length > 3) {
772 } else if (rinfo->prefix_len > 128) {
774 } else if (rinfo->prefix_len > 64) {
775 if (rinfo->length < 2) {
778 } else if (rinfo->prefix_len > 0) {
779 if (rinfo->length < 1) {
784 pref = rinfo->route_pref;
785 if (pref == ICMPV6_ROUTER_PREF_INVALID)
788 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
790 if (rinfo->length == 3)
791 prefix = (struct in6_addr *)rinfo->prefix;
793 /* this function is safe */
794 ipv6_addr_prefix(&prefix_buf,
795 (struct in6_addr *)rinfo->prefix,
797 prefix = &prefix_buf;
800 if (rinfo->prefix_len == 0)
801 rt = rt6_get_dflt_router(gwaddr, dev);
803 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
804 gwaddr, dev->ifindex);
806 if (rt && !lifetime) {
812 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
815 rt->rt6i_flags = RTF_ROUTEINFO |
816 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
819 if (!addrconf_finite_timeout(lifetime))
820 rt6_clean_expires(rt);
822 rt6_set_expires(rt, jiffies + HZ * lifetime);
830 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
831 struct in6_addr *saddr)
833 struct fib6_node *pn;
835 if (fn->fn_flags & RTN_TL_ROOT)
838 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
839 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
842 if (fn->fn_flags & RTN_RTINFO)
847 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
848 struct fib6_table *table,
849 struct flowi6 *fl6, int flags)
851 struct fib6_node *fn;
854 read_lock_bh(&table->tb6_lock);
855 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
858 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
859 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
860 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
861 if (rt == net->ipv6.ip6_null_entry) {
862 fn = fib6_backtrack(fn, &fl6->saddr);
866 dst_use(&rt->dst, jiffies);
867 read_unlock_bh(&table->tb6_lock);
872 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
875 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
877 EXPORT_SYMBOL_GPL(ip6_route_lookup);
879 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
880 const struct in6_addr *saddr, int oif, int strict)
882 struct flowi6 fl6 = {
886 struct dst_entry *dst;
887 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
890 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
891 flags |= RT6_LOOKUP_F_HAS_SADDR;
894 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
896 return (struct rt6_info *) dst;
902 EXPORT_SYMBOL(rt6_lookup);
904 /* ip6_ins_rt is called with FREE table->tb6_lock.
905 It takes new route entry, the addition fails by any reason the
906 route is freed. In any case, if caller does not hold it, it may
910 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
911 struct mx6_config *mxc)
914 struct fib6_table *table;
916 table = rt->rt6i_table;
917 write_lock_bh(&table->tb6_lock);
918 err = fib6_add(&table->tb6_root, rt, info, mxc);
919 write_unlock_bh(&table->tb6_lock);
924 int ip6_ins_rt(struct rt6_info *rt)
926 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
927 struct mx6_config mxc = { .mx = NULL, };
929 return __ip6_ins_rt(rt, &info, &mxc);
932 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
933 const struct in6_addr *daddr,
934 const struct in6_addr *saddr)
942 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
943 ort = (struct rt6_info *)ort->dst.from;
945 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
950 ip6_rt_copy_init(rt, ort);
951 rt->rt6i_flags |= RTF_CACHE;
953 rt->dst.flags |= DST_HOST;
954 rt->rt6i_dst.addr = *daddr;
955 rt->rt6i_dst.plen = 128;
957 if (!rt6_is_gw_or_nonexthop(ort)) {
958 if (ort->rt6i_dst.plen != 128 &&
959 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
960 rt->rt6i_flags |= RTF_ANYCAST;
961 #ifdef CONFIG_IPV6_SUBTREES
962 if (rt->rt6i_src.plen && saddr) {
963 rt->rt6i_src.addr = *saddr;
964 rt->rt6i_src.plen = 128;
972 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
974 struct rt6_info *pcpu_rt;
976 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
977 rt->dst.dev, rt->dst.flags);
981 ip6_rt_copy_init(pcpu_rt, rt);
982 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
983 pcpu_rt->rt6i_flags |= RTF_PCPU;
987 /* It should be called with read_lock_bh(&tb6_lock) acquired */
988 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
990 struct rt6_info *pcpu_rt, **p;
992 p = this_cpu_ptr(rt->rt6i_pcpu);
996 dst_hold(&pcpu_rt->dst);
997 rt6_dst_from_metrics_check(pcpu_rt);
1002 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1004 struct fib6_table *table = rt->rt6i_table;
1005 struct rt6_info *pcpu_rt, *prev, **p;
1007 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1009 struct net *net = dev_net(rt->dst.dev);
1011 dst_hold(&net->ipv6.ip6_null_entry->dst);
1012 return net->ipv6.ip6_null_entry;
1015 read_lock_bh(&table->tb6_lock);
1016 if (rt->rt6i_pcpu) {
1017 p = this_cpu_ptr(rt->rt6i_pcpu);
1018 prev = cmpxchg(p, NULL, pcpu_rt);
1020 /* If someone did it before us, return prev instead */
1021 dst_destroy(&pcpu_rt->dst);
1025 /* rt has been removed from the fib6 tree
1026 * before we have a chance to acquire the read_lock.
1027 * In this case, don't brother to create a pcpu rt
1028 * since rt is going away anyway. The next
1029 * dst_check() will trigger a re-lookup.
1031 dst_destroy(&pcpu_rt->dst);
1034 dst_hold(&pcpu_rt->dst);
1035 rt6_dst_from_metrics_check(pcpu_rt);
1036 read_unlock_bh(&table->tb6_lock);
1040 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1041 struct flowi6 *fl6, int flags)
1043 struct fib6_node *fn, *saved_fn;
1044 struct rt6_info *rt;
1047 strict |= flags & RT6_LOOKUP_F_IFACE;
1048 if (net->ipv6.devconf_all->forwarding == 0)
1049 strict |= RT6_LOOKUP_F_REACHABLE;
1051 read_lock_bh(&table->tb6_lock);
1053 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1056 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1060 rt = rt6_select(fn, oif, strict);
1061 if (rt->rt6i_nsiblings)
1062 rt = rt6_multipath_select(rt, fl6, oif, strict);
1063 if (rt == net->ipv6.ip6_null_entry) {
1064 fn = fib6_backtrack(fn, &fl6->saddr);
1066 goto redo_rt6_select;
1067 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1068 /* also consider unreachable route */
1069 strict &= ~RT6_LOOKUP_F_REACHABLE;
1071 goto redo_rt6_select;
1076 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1077 dst_use(&rt->dst, jiffies);
1078 read_unlock_bh(&table->tb6_lock);
1080 rt6_dst_from_metrics_check(rt);
1082 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1083 !(rt->rt6i_flags & RTF_GATEWAY))) {
1084 /* Create a RTF_CACHE clone which will not be
1085 * owned by the fib6 tree. It is for the special case where
1086 * the daddr in the skb during the neighbor look-up is different
1087 * from the fl6->daddr used to look-up route here.
1090 struct rt6_info *uncached_rt;
1092 dst_use(&rt->dst, jiffies);
1093 read_unlock_bh(&table->tb6_lock);
1095 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1096 dst_release(&rt->dst);
1099 rt6_uncached_list_add(uncached_rt);
1101 uncached_rt = net->ipv6.ip6_null_entry;
1103 dst_hold(&uncached_rt->dst);
1107 /* Get a percpu copy */
1109 struct rt6_info *pcpu_rt;
1111 rt->dst.lastuse = jiffies;
1113 pcpu_rt = rt6_get_pcpu_route(rt);
1116 read_unlock_bh(&table->tb6_lock);
1118 /* We have to do the read_unlock first
1119 * because rt6_make_pcpu_route() may trigger
1120 * ip6_dst_gc() which will take the write_lock.
1123 read_unlock_bh(&table->tb6_lock);
1124 pcpu_rt = rt6_make_pcpu_route(rt);
1125 dst_release(&rt->dst);
1133 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1134 struct flowi6 *fl6, int flags)
1136 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1139 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1140 struct net_device *dev,
1141 struct flowi6 *fl6, int flags)
1143 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1144 flags |= RT6_LOOKUP_F_IFACE;
1146 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1149 void ip6_route_input(struct sk_buff *skb)
1151 const struct ipv6hdr *iph = ipv6_hdr(skb);
1152 struct net *net = dev_net(skb->dev);
1153 int flags = RT6_LOOKUP_F_HAS_SADDR;
1154 struct ip_tunnel_info *tun_info;
1155 struct flowi6 fl6 = {
1156 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1157 .daddr = iph->daddr,
1158 .saddr = iph->saddr,
1159 .flowlabel = ip6_flowinfo(iph),
1160 .flowi6_mark = skb->mark,
1161 .flowi6_proto = iph->nexthdr,
1164 tun_info = skb_tunnel_info(skb);
1165 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1166 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1168 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1171 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1172 struct flowi6 *fl6, int flags)
1174 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1177 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1178 struct flowi6 *fl6, int flags)
1180 struct dst_entry *dst;
1183 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1187 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1189 any_src = ipv6_addr_any(&fl6->saddr);
1190 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1191 (fl6->flowi6_oif && any_src))
1192 flags |= RT6_LOOKUP_F_IFACE;
1195 flags |= RT6_LOOKUP_F_HAS_SADDR;
1197 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1199 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1201 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1203 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1205 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1206 struct dst_entry *new = NULL;
1208 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1214 new->input = dst_discard;
1215 new->output = dst_discard_out;
1217 dst_copy_metrics(new, &ort->dst);
1218 rt->rt6i_idev = ort->rt6i_idev;
1220 in6_dev_hold(rt->rt6i_idev);
1222 rt->rt6i_gateway = ort->rt6i_gateway;
1223 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1224 rt->rt6i_metric = 0;
1226 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1227 #ifdef CONFIG_IPV6_SUBTREES
1228 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1234 dst_release(dst_orig);
1235 return new ? new : ERR_PTR(-ENOMEM);
1239 * Destination cache support functions
1242 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1245 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1246 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1249 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1251 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1254 if (rt6_check_expired(rt))
1260 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1262 if (!__rt6_check_expired(rt) &&
1263 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1264 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1270 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1272 struct rt6_info *rt;
1274 rt = (struct rt6_info *) dst;
1276 /* All IPV6 dsts are created with ->obsolete set to the value
1277 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1278 * into this function always.
1281 rt6_dst_from_metrics_check(rt);
1283 if (rt->rt6i_flags & RTF_PCPU ||
1284 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1285 return rt6_dst_from_check(rt, cookie);
1287 return rt6_check(rt, cookie);
1290 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1292 struct rt6_info *rt = (struct rt6_info *) dst;
1295 if (rt->rt6i_flags & RTF_CACHE) {
1296 if (rt6_check_expired(rt)) {
1308 static void ip6_link_failure(struct sk_buff *skb)
1310 struct rt6_info *rt;
1312 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1314 rt = (struct rt6_info *) skb_dst(skb);
1316 if (rt->rt6i_flags & RTF_CACHE) {
1319 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1320 rt->rt6i_node->fn_sernum = -1;
1325 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1327 struct net *net = dev_net(rt->dst.dev);
1329 rt->rt6i_flags |= RTF_MODIFIED;
1330 rt->rt6i_pmtu = mtu;
1331 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1334 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1336 return !(rt->rt6i_flags & RTF_CACHE) &&
1337 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1340 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1341 const struct ipv6hdr *iph, u32 mtu)
1343 struct rt6_info *rt6 = (struct rt6_info *)dst;
1345 if (rt6->rt6i_flags & RTF_LOCAL)
1349 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1350 if (mtu >= dst_mtu(dst))
1353 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1354 rt6_do_update_pmtu(rt6, mtu);
1356 const struct in6_addr *daddr, *saddr;
1357 struct rt6_info *nrt6;
1360 daddr = &iph->daddr;
1361 saddr = &iph->saddr;
1363 daddr = &sk->sk_v6_daddr;
1364 saddr = &inet6_sk(sk)->saddr;
1368 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1370 rt6_do_update_pmtu(nrt6, mtu);
1372 /* ip6_ins_rt(nrt6) will bump the
1373 * rt6->rt6i_node->fn_sernum
1374 * which will fail the next rt6_check() and
1375 * invalidate the sk->sk_dst_cache.
1382 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1383 struct sk_buff *skb, u32 mtu)
1385 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1388 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1391 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1392 struct dst_entry *dst;
1395 memset(&fl6, 0, sizeof(fl6));
1396 fl6.flowi6_oif = oif;
1397 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1398 fl6.daddr = iph->daddr;
1399 fl6.saddr = iph->saddr;
1400 fl6.flowlabel = ip6_flowinfo(iph);
1402 dst = ip6_route_output(net, NULL, &fl6);
1404 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1407 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1409 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1411 ip6_update_pmtu(skb, sock_net(sk), mtu,
1412 sk->sk_bound_dev_if, sk->sk_mark);
1414 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1416 /* Handle redirects */
1417 struct ip6rd_flowi {
1419 struct in6_addr gateway;
1422 static struct rt6_info *__ip6_route_redirect(struct net *net,
1423 struct fib6_table *table,
1427 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1428 struct rt6_info *rt;
1429 struct fib6_node *fn;
1431 /* Get the "current" route for this destination and
1432 * check if the redirect has come from approriate router.
1434 * RFC 4861 specifies that redirects should only be
1435 * accepted if they come from the nexthop to the target.
1436 * Due to the way the routes are chosen, this notion
1437 * is a bit fuzzy and one might need to check all possible
1441 read_lock_bh(&table->tb6_lock);
1442 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1444 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1445 if (rt6_check_expired(rt))
1449 if (!(rt->rt6i_flags & RTF_GATEWAY))
1451 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1453 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1459 rt = net->ipv6.ip6_null_entry;
1460 else if (rt->dst.error) {
1461 rt = net->ipv6.ip6_null_entry;
1465 if (rt == net->ipv6.ip6_null_entry) {
1466 fn = fib6_backtrack(fn, &fl6->saddr);
1474 read_unlock_bh(&table->tb6_lock);
1479 static struct dst_entry *ip6_route_redirect(struct net *net,
1480 const struct flowi6 *fl6,
1481 const struct in6_addr *gateway)
1483 int flags = RT6_LOOKUP_F_HAS_SADDR;
1484 struct ip6rd_flowi rdfl;
1487 rdfl.gateway = *gateway;
1489 return fib6_rule_lookup(net, &rdfl.fl6,
1490 flags, __ip6_route_redirect);
1493 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1495 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1496 struct dst_entry *dst;
1499 memset(&fl6, 0, sizeof(fl6));
1500 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1501 fl6.flowi6_oif = oif;
1502 fl6.flowi6_mark = mark;
1503 fl6.daddr = iph->daddr;
1504 fl6.saddr = iph->saddr;
1505 fl6.flowlabel = ip6_flowinfo(iph);
1507 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1508 rt6_do_redirect(dst, NULL, skb);
1511 EXPORT_SYMBOL_GPL(ip6_redirect);
1513 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1516 const struct ipv6hdr *iph = ipv6_hdr(skb);
1517 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1518 struct dst_entry *dst;
1521 memset(&fl6, 0, sizeof(fl6));
1522 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1523 fl6.flowi6_oif = oif;
1524 fl6.flowi6_mark = mark;
1525 fl6.daddr = msg->dest;
1526 fl6.saddr = iph->daddr;
1528 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1529 rt6_do_redirect(dst, NULL, skb);
1533 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1535 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1537 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1539 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1541 struct net_device *dev = dst->dev;
1542 unsigned int mtu = dst_mtu(dst);
1543 struct net *net = dev_net(dev);
1545 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1547 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1548 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1551 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1552 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1553 * IPV6_MAXPLEN is also valid and means: "any MSS,
1554 * rely only on pmtu discovery"
1556 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1561 static unsigned int ip6_mtu(const struct dst_entry *dst)
1563 const struct rt6_info *rt = (const struct rt6_info *)dst;
1564 unsigned int mtu = rt->rt6i_pmtu;
1565 struct inet6_dev *idev;
1570 mtu = dst_metric_raw(dst, RTAX_MTU);
1577 idev = __in6_dev_get(dst->dev);
1579 mtu = idev->cnf.mtu6;
1583 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1586 static struct dst_entry *icmp6_dst_gc_list;
1587 static DEFINE_SPINLOCK(icmp6_dst_lock);
1589 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1592 struct dst_entry *dst;
1593 struct rt6_info *rt;
1594 struct inet6_dev *idev = in6_dev_get(dev);
1595 struct net *net = dev_net(dev);
1597 if (unlikely(!idev))
1598 return ERR_PTR(-ENODEV);
1600 rt = ip6_dst_alloc(net, dev, 0);
1601 if (unlikely(!rt)) {
1603 dst = ERR_PTR(-ENOMEM);
1607 rt->dst.flags |= DST_HOST;
1608 rt->dst.output = ip6_output;
1609 atomic_set(&rt->dst.__refcnt, 1);
1610 rt->rt6i_gateway = fl6->daddr;
1611 rt->rt6i_dst.addr = fl6->daddr;
1612 rt->rt6i_dst.plen = 128;
1613 rt->rt6i_idev = idev;
1614 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1616 spin_lock_bh(&icmp6_dst_lock);
1617 rt->dst.next = icmp6_dst_gc_list;
1618 icmp6_dst_gc_list = &rt->dst;
1619 spin_unlock_bh(&icmp6_dst_lock);
1621 fib6_force_start_gc(net);
1623 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1629 int icmp6_dst_gc(void)
1631 struct dst_entry *dst, **pprev;
1634 spin_lock_bh(&icmp6_dst_lock);
1635 pprev = &icmp6_dst_gc_list;
1637 while ((dst = *pprev) != NULL) {
1638 if (!atomic_read(&dst->__refcnt)) {
1647 spin_unlock_bh(&icmp6_dst_lock);
1652 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1655 struct dst_entry *dst, **pprev;
1657 spin_lock_bh(&icmp6_dst_lock);
1658 pprev = &icmp6_dst_gc_list;
1659 while ((dst = *pprev) != NULL) {
1660 struct rt6_info *rt = (struct rt6_info *) dst;
1661 if (func(rt, arg)) {
1668 spin_unlock_bh(&icmp6_dst_lock);
1671 static int ip6_dst_gc(struct dst_ops *ops)
1673 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1674 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1675 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1676 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1677 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1678 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1681 entries = dst_entries_get_fast(ops);
1682 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1683 entries <= rt_max_size)
1686 net->ipv6.ip6_rt_gc_expire++;
1687 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1688 entries = dst_entries_get_slow(ops);
1689 if (entries < ops->gc_thresh)
1690 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1692 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1693 return entries > rt_max_size;
1696 static int ip6_convert_metrics(struct mx6_config *mxc,
1697 const struct fib6_config *cfg)
1699 bool ecn_ca = false;
1707 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1711 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1712 int type = nla_type(nla);
1717 if (unlikely(type > RTAX_MAX))
1720 if (type == RTAX_CC_ALGO) {
1721 char tmp[TCP_CA_NAME_MAX];
1723 nla_strlcpy(tmp, nla, sizeof(tmp));
1724 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1725 if (val == TCP_CA_UNSPEC)
1728 val = nla_get_u32(nla);
1730 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1734 __set_bit(type - 1, mxc->mx_valid);
1738 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1739 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1749 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1751 struct net *net = cfg->fc_nlinfo.nl_net;
1752 struct rt6_info *rt = NULL;
1753 struct net_device *dev = NULL;
1754 struct inet6_dev *idev = NULL;
1755 struct fib6_table *table;
1759 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1761 #ifndef CONFIG_IPV6_SUBTREES
1762 if (cfg->fc_src_len)
1765 if (cfg->fc_ifindex) {
1767 dev = dev_get_by_index(net, cfg->fc_ifindex);
1770 idev = in6_dev_get(dev);
1775 if (cfg->fc_metric == 0)
1776 cfg->fc_metric = IP6_RT_PRIO_USER;
1779 if (cfg->fc_nlinfo.nlh &&
1780 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1781 table = fib6_get_table(net, cfg->fc_table);
1783 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1784 table = fib6_new_table(net, cfg->fc_table);
1787 table = fib6_new_table(net, cfg->fc_table);
1793 rt = ip6_dst_alloc(net, NULL,
1794 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1801 if (cfg->fc_flags & RTF_EXPIRES)
1802 rt6_set_expires(rt, jiffies +
1803 clock_t_to_jiffies(cfg->fc_expires));
1805 rt6_clean_expires(rt);
1807 if (cfg->fc_protocol == RTPROT_UNSPEC)
1808 cfg->fc_protocol = RTPROT_BOOT;
1809 rt->rt6i_protocol = cfg->fc_protocol;
1811 addr_type = ipv6_addr_type(&cfg->fc_dst);
1813 if (addr_type & IPV6_ADDR_MULTICAST)
1814 rt->dst.input = ip6_mc_input;
1815 else if (cfg->fc_flags & RTF_LOCAL)
1816 rt->dst.input = ip6_input;
1818 rt->dst.input = ip6_forward;
1820 rt->dst.output = ip6_output;
1822 if (cfg->fc_encap) {
1823 struct lwtunnel_state *lwtstate;
1825 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1826 cfg->fc_encap, AF_INET6, cfg,
1830 rt->dst.lwtstate = lwtstate_get(lwtstate);
1831 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1832 rt->dst.lwtstate->orig_output = rt->dst.output;
1833 rt->dst.output = lwtunnel_output;
1835 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1836 rt->dst.lwtstate->orig_input = rt->dst.input;
1837 rt->dst.input = lwtunnel_input;
1841 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1842 rt->rt6i_dst.plen = cfg->fc_dst_len;
1843 if (rt->rt6i_dst.plen == 128)
1844 rt->dst.flags |= DST_HOST;
1846 #ifdef CONFIG_IPV6_SUBTREES
1847 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1848 rt->rt6i_src.plen = cfg->fc_src_len;
1851 rt->rt6i_metric = cfg->fc_metric;
1853 /* We cannot add true routes via loopback here,
1854 they would result in kernel looping; promote them to reject routes
1856 if ((cfg->fc_flags & RTF_REJECT) ||
1857 (dev && (dev->flags & IFF_LOOPBACK) &&
1858 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1859 !(cfg->fc_flags & RTF_LOCAL))) {
1860 /* hold loopback dev/idev if we haven't done so. */
1861 if (dev != net->loopback_dev) {
1866 dev = net->loopback_dev;
1868 idev = in6_dev_get(dev);
1874 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1875 switch (cfg->fc_type) {
1877 rt->dst.error = -EINVAL;
1878 rt->dst.output = dst_discard_out;
1879 rt->dst.input = dst_discard;
1882 rt->dst.error = -EACCES;
1883 rt->dst.output = ip6_pkt_prohibit_out;
1884 rt->dst.input = ip6_pkt_prohibit;
1887 case RTN_UNREACHABLE:
1889 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1890 : (cfg->fc_type == RTN_UNREACHABLE)
1891 ? -EHOSTUNREACH : -ENETUNREACH;
1892 rt->dst.output = ip6_pkt_discard_out;
1893 rt->dst.input = ip6_pkt_discard;
1899 if (cfg->fc_flags & RTF_GATEWAY) {
1900 const struct in6_addr *gw_addr;
1903 gw_addr = &cfg->fc_gateway;
1904 gwa_type = ipv6_addr_type(gw_addr);
1906 /* if gw_addr is local we will fail to detect this in case
1907 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1908 * will return already-added prefix route via interface that
1909 * prefix route was assigned to, which might be non-loopback.
1912 if (ipv6_chk_addr_and_flags(net, gw_addr,
1913 gwa_type & IPV6_ADDR_LINKLOCAL ?
1917 rt->rt6i_gateway = *gw_addr;
1919 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1920 struct rt6_info *grt;
1922 /* IPv6 strictly inhibits using not link-local
1923 addresses as nexthop address.
1924 Otherwise, router will not able to send redirects.
1925 It is very good, but in some (rare!) circumstances
1926 (SIT, PtP, NBMA NOARP links) it is handy to allow
1927 some exceptions. --ANK
1929 if (!(gwa_type & IPV6_ADDR_UNICAST))
1932 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1934 err = -EHOSTUNREACH;
1938 if (dev != grt->dst.dev) {
1944 idev = grt->rt6i_idev;
1946 in6_dev_hold(grt->rt6i_idev);
1948 if (!(grt->rt6i_flags & RTF_GATEWAY))
1956 if (!dev || (dev->flags & IFF_LOOPBACK))
1964 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1965 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1969 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1970 rt->rt6i_prefsrc.plen = 128;
1972 rt->rt6i_prefsrc.plen = 0;
1974 rt->rt6i_flags = cfg->fc_flags;
1978 rt->rt6i_idev = idev;
1979 rt->rt6i_table = table;
1981 cfg->fc_nlinfo.nl_net = dev_net(dev);
1992 return ERR_PTR(err);
1995 int ip6_route_add(struct fib6_config *cfg)
1997 struct mx6_config mxc = { .mx = NULL, };
1998 struct rt6_info *rt;
2001 rt = ip6_route_info_create(cfg);
2008 err = ip6_convert_metrics(&mxc, cfg);
2012 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2024 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2027 struct fib6_table *table;
2028 struct net *net = dev_net(rt->dst.dev);
2030 if (rt == net->ipv6.ip6_null_entry ||
2031 rt->dst.flags & DST_NOCACHE) {
2036 table = rt->rt6i_table;
2037 write_lock_bh(&table->tb6_lock);
2038 err = fib6_del(rt, info);
2039 write_unlock_bh(&table->tb6_lock);
2046 int ip6_del_rt(struct rt6_info *rt)
2048 struct nl_info info = {
2049 .nl_net = dev_net(rt->dst.dev),
2051 return __ip6_del_rt(rt, &info);
2054 static int ip6_route_del(struct fib6_config *cfg)
2056 struct fib6_table *table;
2057 struct fib6_node *fn;
2058 struct rt6_info *rt;
2061 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2065 read_lock_bh(&table->tb6_lock);
2067 fn = fib6_locate(&table->tb6_root,
2068 &cfg->fc_dst, cfg->fc_dst_len,
2069 &cfg->fc_src, cfg->fc_src_len);
2072 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2073 if ((rt->rt6i_flags & RTF_CACHE) &&
2074 !(cfg->fc_flags & RTF_CACHE))
2076 if (cfg->fc_ifindex &&
2078 rt->dst.dev->ifindex != cfg->fc_ifindex))
2080 if (cfg->fc_flags & RTF_GATEWAY &&
2081 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2083 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2086 read_unlock_bh(&table->tb6_lock);
2088 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2091 read_unlock_bh(&table->tb6_lock);
2096 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2098 struct netevent_redirect netevent;
2099 struct rt6_info *rt, *nrt = NULL;
2100 struct ndisc_options ndopts;
2101 struct inet6_dev *in6_dev;
2102 struct neighbour *neigh;
2104 int optlen, on_link;
2107 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2108 optlen -= sizeof(*msg);
2111 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2115 msg = (struct rd_msg *)icmp6_hdr(skb);
2117 if (ipv6_addr_is_multicast(&msg->dest)) {
2118 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2123 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2125 } else if (ipv6_addr_type(&msg->target) !=
2126 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2127 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2131 in6_dev = __in6_dev_get(skb->dev);
2134 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2138 * The IP source address of the Redirect MUST be the same as the current
2139 * first-hop router for the specified ICMP Destination Address.
2142 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2143 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2148 if (ndopts.nd_opts_tgt_lladdr) {
2149 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2152 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2157 rt = (struct rt6_info *) dst;
2158 if (rt->rt6i_flags & RTF_REJECT) {
2159 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2163 /* Redirect received -> path was valid.
2164 * Look, redirects are sent only in response to data packets,
2165 * so that this nexthop apparently is reachable. --ANK
2167 dst_confirm(&rt->dst);
2169 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2174 * We have finally decided to accept it.
2177 neigh_update(neigh, lladdr, NUD_STALE,
2178 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2179 NEIGH_UPDATE_F_OVERRIDE|
2180 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2181 NEIGH_UPDATE_F_ISROUTER))
2184 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2188 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2190 nrt->rt6i_flags &= ~RTF_GATEWAY;
2192 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2194 if (ip6_ins_rt(nrt))
2197 netevent.old = &rt->dst;
2198 netevent.new = &nrt->dst;
2199 netevent.daddr = &msg->dest;
2200 netevent.neigh = neigh;
2201 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2203 if (rt->rt6i_flags & RTF_CACHE) {
2204 rt = (struct rt6_info *) dst_clone(&rt->dst);
2209 neigh_release(neigh);
2213 * Misc support functions
2216 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2218 BUG_ON(from->dst.from);
2220 rt->rt6i_flags &= ~RTF_EXPIRES;
2221 dst_hold(&from->dst);
2222 rt->dst.from = &from->dst;
2223 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2226 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2228 rt->dst.input = ort->dst.input;
2229 rt->dst.output = ort->dst.output;
2230 rt->rt6i_dst = ort->rt6i_dst;
2231 rt->dst.error = ort->dst.error;
2232 rt->rt6i_idev = ort->rt6i_idev;
2234 in6_dev_hold(rt->rt6i_idev);
2235 rt->dst.lastuse = jiffies;
2236 rt->rt6i_gateway = ort->rt6i_gateway;
2237 rt->rt6i_flags = ort->rt6i_flags;
2238 rt6_set_from(rt, ort);
2239 rt->rt6i_metric = ort->rt6i_metric;
2240 #ifdef CONFIG_IPV6_SUBTREES
2241 rt->rt6i_src = ort->rt6i_src;
2243 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2244 rt->rt6i_table = ort->rt6i_table;
2245 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2248 #ifdef CONFIG_IPV6_ROUTE_INFO
2249 static struct rt6_info *rt6_get_route_info(struct net *net,
2250 const struct in6_addr *prefix, int prefixlen,
2251 const struct in6_addr *gwaddr, int ifindex)
2253 struct fib6_node *fn;
2254 struct rt6_info *rt = NULL;
2255 struct fib6_table *table;
2257 table = fib6_get_table(net, RT6_TABLE_INFO);
2261 read_lock_bh(&table->tb6_lock);
2262 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2266 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2267 if (rt->dst.dev->ifindex != ifindex)
2269 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2271 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2277 read_unlock_bh(&table->tb6_lock);
2281 static struct rt6_info *rt6_add_route_info(struct net *net,
2282 const struct in6_addr *prefix, int prefixlen,
2283 const struct in6_addr *gwaddr, int ifindex,
2286 struct fib6_config cfg = {
2287 .fc_metric = IP6_RT_PRIO_USER,
2288 .fc_ifindex = ifindex,
2289 .fc_dst_len = prefixlen,
2290 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2291 RTF_UP | RTF_PREF(pref),
2292 .fc_nlinfo.portid = 0,
2293 .fc_nlinfo.nlh = NULL,
2294 .fc_nlinfo.nl_net = net,
2297 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2298 cfg.fc_dst = *prefix;
2299 cfg.fc_gateway = *gwaddr;
2301 /* We should treat it as a default route if prefix length is 0. */
2303 cfg.fc_flags |= RTF_DEFAULT;
2305 ip6_route_add(&cfg);
2307 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2311 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2313 struct rt6_info *rt;
2314 struct fib6_table *table;
2316 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2320 read_lock_bh(&table->tb6_lock);
2321 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2322 if (dev == rt->dst.dev &&
2323 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2324 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2329 read_unlock_bh(&table->tb6_lock);
2333 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2334 struct net_device *dev,
2337 struct fib6_config cfg = {
2338 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2339 .fc_metric = IP6_RT_PRIO_USER,
2340 .fc_ifindex = dev->ifindex,
2341 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2342 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2343 .fc_nlinfo.portid = 0,
2344 .fc_nlinfo.nlh = NULL,
2345 .fc_nlinfo.nl_net = dev_net(dev),
2348 cfg.fc_gateway = *gwaddr;
2350 ip6_route_add(&cfg);
2352 return rt6_get_dflt_router(gwaddr, dev);
2355 void rt6_purge_dflt_routers(struct net *net)
2357 struct rt6_info *rt;
2358 struct fib6_table *table;
2360 /* NOTE: Keep consistent with rt6_get_dflt_router */
2361 table = fib6_get_table(net, RT6_TABLE_DFLT);
2366 read_lock_bh(&table->tb6_lock);
2367 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2368 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2369 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2371 read_unlock_bh(&table->tb6_lock);
2376 read_unlock_bh(&table->tb6_lock);
2379 static void rtmsg_to_fib6_config(struct net *net,
2380 struct in6_rtmsg *rtmsg,
2381 struct fib6_config *cfg)
2383 memset(cfg, 0, sizeof(*cfg));
2385 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2387 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2388 cfg->fc_metric = rtmsg->rtmsg_metric;
2389 cfg->fc_expires = rtmsg->rtmsg_info;
2390 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2391 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2392 cfg->fc_flags = rtmsg->rtmsg_flags;
2394 cfg->fc_nlinfo.nl_net = net;
2396 cfg->fc_dst = rtmsg->rtmsg_dst;
2397 cfg->fc_src = rtmsg->rtmsg_src;
2398 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2401 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2403 struct fib6_config cfg;
2404 struct in6_rtmsg rtmsg;
2408 case SIOCADDRT: /* Add a route */
2409 case SIOCDELRT: /* Delete a route */
2410 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2412 err = copy_from_user(&rtmsg, arg,
2413 sizeof(struct in6_rtmsg));
2417 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2422 err = ip6_route_add(&cfg);
2425 err = ip6_route_del(&cfg);
2439 * Drop the packet on the floor
2442 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2445 struct dst_entry *dst = skb_dst(skb);
2446 switch (ipstats_mib_noroutes) {
2447 case IPSTATS_MIB_INNOROUTES:
2448 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2449 if (type == IPV6_ADDR_ANY) {
2450 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2451 IPSTATS_MIB_INADDRERRORS);
2455 case IPSTATS_MIB_OUTNOROUTES:
2456 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2457 ipstats_mib_noroutes);
2460 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2465 static int ip6_pkt_discard(struct sk_buff *skb)
2467 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2470 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2472 skb->dev = skb_dst(skb)->dev;
2473 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2476 static int ip6_pkt_prohibit(struct sk_buff *skb)
2478 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2481 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2483 skb->dev = skb_dst(skb)->dev;
2484 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2488 * Allocate a dst for local (unicast / anycast) address.
2491 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2492 const struct in6_addr *addr,
2496 struct net *net = dev_net(idev->dev);
2497 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2500 return ERR_PTR(-ENOMEM);
2504 rt->dst.flags |= DST_HOST;
2505 rt->dst.input = ip6_input;
2506 rt->dst.output = ip6_output;
2507 rt->rt6i_idev = idev;
2509 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2511 rt->rt6i_flags |= RTF_ANYCAST;
2513 rt->rt6i_flags |= RTF_LOCAL;
2515 rt->rt6i_gateway = *addr;
2516 rt->rt6i_dst.addr = *addr;
2517 rt->rt6i_dst.plen = 128;
2518 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2519 rt->rt6i_table = fib6_get_table(net, tb_id);
2520 rt->dst.flags |= DST_NOCACHE;
2522 atomic_set(&rt->dst.__refcnt, 1);
2527 int ip6_route_get_saddr(struct net *net,
2528 struct rt6_info *rt,
2529 const struct in6_addr *daddr,
2531 struct in6_addr *saddr)
2533 struct inet6_dev *idev =
2534 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2536 if (rt && rt->rt6i_prefsrc.plen)
2537 *saddr = rt->rt6i_prefsrc.addr;
2539 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2540 daddr, prefs, saddr);
2544 /* remove deleted ip from prefsrc entries */
2545 struct arg_dev_net_ip {
2546 struct net_device *dev;
2548 struct in6_addr *addr;
2551 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2553 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2554 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2555 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2557 if (((void *)rt->dst.dev == dev || !dev) &&
2558 rt != net->ipv6.ip6_null_entry &&
2559 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2560 /* remove prefsrc entry */
2561 rt->rt6i_prefsrc.plen = 0;
2566 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2568 struct net *net = dev_net(ifp->idev->dev);
2569 struct arg_dev_net_ip adni = {
2570 .dev = ifp->idev->dev,
2574 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2577 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2578 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2580 /* Remove routers and update dst entries when gateway turn into host. */
2581 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2583 struct in6_addr *gateway = (struct in6_addr *)arg;
2585 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2586 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2587 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2593 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2595 fib6_clean_all(net, fib6_clean_tohost, gateway);
2598 struct arg_dev_net {
2599 struct net_device *dev;
2603 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2605 const struct arg_dev_net *adn = arg;
2606 const struct net_device *dev = adn->dev;
2608 if ((rt->dst.dev == dev || !dev) &&
2609 rt != adn->net->ipv6.ip6_null_entry)
2615 void rt6_ifdown(struct net *net, struct net_device *dev)
2617 struct arg_dev_net adn = {
2622 fib6_clean_all(net, fib6_ifdown, &adn);
2623 icmp6_clean_all(fib6_ifdown, &adn);
2625 rt6_uncached_list_flush_dev(net, dev);
2628 struct rt6_mtu_change_arg {
2629 struct net_device *dev;
2633 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2635 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2636 struct inet6_dev *idev;
2638 /* In IPv6 pmtu discovery is not optional,
2639 so that RTAX_MTU lock cannot disable it.
2640 We still use this lock to block changes
2641 caused by addrconf/ndisc.
2644 idev = __in6_dev_get(arg->dev);
2648 /* For administrative MTU increase, there is no way to discover
2649 IPv6 PMTU increase, so PMTU increase should be updated here.
2650 Since RFC 1981 doesn't include administrative MTU increase
2651 update PMTU increase is a MUST. (i.e. jumbo frame)
2654 If new MTU is less than route PMTU, this new MTU will be the
2655 lowest MTU in the path, update the route PMTU to reflect PMTU
2656 decreases; if new MTU is greater than route PMTU, and the
2657 old MTU is the lowest MTU in the path, update the route PMTU
2658 to reflect the increase. In this case if the other nodes' MTU
2659 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2662 if (rt->dst.dev == arg->dev &&
2663 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2664 if (rt->rt6i_flags & RTF_CACHE) {
2665 /* For RTF_CACHE with rt6i_pmtu == 0
2666 * (i.e. a redirected route),
2667 * the metrics of its rt->dst.from has already
2670 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2671 rt->rt6i_pmtu = arg->mtu;
2672 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2673 (dst_mtu(&rt->dst) < arg->mtu &&
2674 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2675 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2681 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2683 struct rt6_mtu_change_arg arg = {
2688 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2691 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2692 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2693 [RTA_OIF] = { .type = NLA_U32 },
2694 [RTA_IIF] = { .type = NLA_U32 },
2695 [RTA_PRIORITY] = { .type = NLA_U32 },
2696 [RTA_METRICS] = { .type = NLA_NESTED },
2697 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2698 [RTA_PREF] = { .type = NLA_U8 },
2699 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2700 [RTA_ENCAP] = { .type = NLA_NESTED },
2703 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2704 struct fib6_config *cfg)
2707 struct nlattr *tb[RTA_MAX+1];
2711 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2716 rtm = nlmsg_data(nlh);
2717 memset(cfg, 0, sizeof(*cfg));
2719 cfg->fc_table = rtm->rtm_table;
2720 cfg->fc_dst_len = rtm->rtm_dst_len;
2721 cfg->fc_src_len = rtm->rtm_src_len;
2722 cfg->fc_flags = RTF_UP;
2723 cfg->fc_protocol = rtm->rtm_protocol;
2724 cfg->fc_type = rtm->rtm_type;
2726 if (rtm->rtm_type == RTN_UNREACHABLE ||
2727 rtm->rtm_type == RTN_BLACKHOLE ||
2728 rtm->rtm_type == RTN_PROHIBIT ||
2729 rtm->rtm_type == RTN_THROW)
2730 cfg->fc_flags |= RTF_REJECT;
2732 if (rtm->rtm_type == RTN_LOCAL)
2733 cfg->fc_flags |= RTF_LOCAL;
2735 if (rtm->rtm_flags & RTM_F_CLONED)
2736 cfg->fc_flags |= RTF_CACHE;
2738 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2739 cfg->fc_nlinfo.nlh = nlh;
2740 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2742 if (tb[RTA_GATEWAY]) {
2743 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2744 cfg->fc_flags |= RTF_GATEWAY;
2748 int plen = (rtm->rtm_dst_len + 7) >> 3;
2750 if (nla_len(tb[RTA_DST]) < plen)
2753 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2757 int plen = (rtm->rtm_src_len + 7) >> 3;
2759 if (nla_len(tb[RTA_SRC]) < plen)
2762 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2765 if (tb[RTA_PREFSRC])
2766 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2769 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2771 if (tb[RTA_PRIORITY])
2772 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2774 if (tb[RTA_METRICS]) {
2775 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2776 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2780 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2782 if (tb[RTA_MULTIPATH]) {
2783 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2784 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2788 pref = nla_get_u8(tb[RTA_PREF]);
2789 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2790 pref != ICMPV6_ROUTER_PREF_HIGH)
2791 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2792 cfg->fc_flags |= RTF_PREF(pref);
2796 cfg->fc_encap = tb[RTA_ENCAP];
2798 if (tb[RTA_ENCAP_TYPE])
2799 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2807 struct rt6_info *rt6_info;
2808 struct fib6_config r_cfg;
2809 struct mx6_config mxc;
2810 struct list_head next;
2813 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2817 list_for_each_entry(nh, rt6_nh_list, next) {
2818 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2819 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2820 nh->r_cfg.fc_ifindex);
2824 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2825 struct rt6_info *rt, struct fib6_config *r_cfg)
2828 struct rt6_info *rtnh;
2831 list_for_each_entry(nh, rt6_nh_list, next) {
2832 /* check if rt6_info already exists */
2833 rtnh = nh->rt6_info;
2835 if (rtnh->dst.dev == rt->dst.dev &&
2836 rtnh->rt6i_idev == rt->rt6i_idev &&
2837 ipv6_addr_equal(&rtnh->rt6i_gateway,
2842 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2846 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2851 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2852 list_add_tail(&nh->next, rt6_nh_list);
2857 static int ip6_route_multipath_add(struct fib6_config *cfg)
2859 struct fib6_config r_cfg;
2860 struct rtnexthop *rtnh;
2861 struct rt6_info *rt;
2862 struct rt6_nh *err_nh;
2863 struct rt6_nh *nh, *nh_safe;
2868 int replace = (cfg->fc_nlinfo.nlh &&
2869 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2870 LIST_HEAD(rt6_nh_list);
2872 remaining = cfg->fc_mp_len;
2873 rtnh = (struct rtnexthop *)cfg->fc_mp;
2875 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2876 * rt6_info structs per nexthop
2878 while (rtnh_ok(rtnh, remaining)) {
2879 memcpy(&r_cfg, cfg, sizeof(*cfg));
2880 if (rtnh->rtnh_ifindex)
2881 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2883 attrlen = rtnh_attrlen(rtnh);
2885 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2887 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2889 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2890 r_cfg.fc_flags |= RTF_GATEWAY;
2892 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2893 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2895 r_cfg.fc_encap_type = nla_get_u16(nla);
2898 rt = ip6_route_info_create(&r_cfg);
2905 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2911 rtnh = rtnh_next(rtnh, &remaining);
2915 list_for_each_entry(nh, &rt6_nh_list, next) {
2916 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2917 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2918 nh->rt6_info = NULL;
2921 ip6_print_replace_route_err(&rt6_nh_list);
2926 /* Because each route is added like a single route we remove
2927 * these flags after the first nexthop: if there is a collision,
2928 * we have already failed to add the first nexthop:
2929 * fib6_add_rt2node() has rejected it; when replacing, old
2930 * nexthops have been replaced by first new, the rest should
2933 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2941 /* Delete routes that were already added */
2942 list_for_each_entry(nh, &rt6_nh_list, next) {
2945 ip6_route_del(&nh->r_cfg);
2949 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2951 dst_free(&nh->rt6_info->dst);
2953 list_del(&nh->next);
2960 static int ip6_route_multipath_del(struct fib6_config *cfg)
2962 struct fib6_config r_cfg;
2963 struct rtnexthop *rtnh;
2966 int err = 1, last_err = 0;
2968 remaining = cfg->fc_mp_len;
2969 rtnh = (struct rtnexthop *)cfg->fc_mp;
2971 /* Parse a Multipath Entry */
2972 while (rtnh_ok(rtnh, remaining)) {
2973 memcpy(&r_cfg, cfg, sizeof(*cfg));
2974 if (rtnh->rtnh_ifindex)
2975 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2977 attrlen = rtnh_attrlen(rtnh);
2979 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2981 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2983 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2984 r_cfg.fc_flags |= RTF_GATEWAY;
2987 err = ip6_route_del(&r_cfg);
2991 rtnh = rtnh_next(rtnh, &remaining);
2997 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2999 struct fib6_config cfg;
3002 err = rtm_to_fib6_config(skb, nlh, &cfg);
3007 return ip6_route_multipath_del(&cfg);
3009 return ip6_route_del(&cfg);
3012 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3014 struct fib6_config cfg;
3017 err = rtm_to_fib6_config(skb, nlh, &cfg);
3022 return ip6_route_multipath_add(&cfg);
3024 return ip6_route_add(&cfg);
3027 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3029 return NLMSG_ALIGN(sizeof(struct rtmsg))
3030 + nla_total_size(16) /* RTA_SRC */
3031 + nla_total_size(16) /* RTA_DST */
3032 + nla_total_size(16) /* RTA_GATEWAY */
3033 + nla_total_size(16) /* RTA_PREFSRC */
3034 + nla_total_size(4) /* RTA_TABLE */
3035 + nla_total_size(4) /* RTA_IIF */
3036 + nla_total_size(4) /* RTA_OIF */
3037 + nla_total_size(4) /* RTA_PRIORITY */
3038 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3039 + nla_total_size(sizeof(struct rta_cacheinfo))
3040 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3041 + nla_total_size(1) /* RTA_PREF */
3042 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3045 static int rt6_fill_node(struct net *net,
3046 struct sk_buff *skb, struct rt6_info *rt,
3047 struct in6_addr *dst, struct in6_addr *src,
3048 int iif, int type, u32 portid, u32 seq,
3049 int prefix, int nowait, unsigned int flags)
3051 u32 metrics[RTAX_MAX];
3053 struct nlmsghdr *nlh;
3057 if (prefix) { /* user wants prefix routes only */
3058 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3059 /* success since this is not a prefix route */
3064 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3068 rtm = nlmsg_data(nlh);
3069 rtm->rtm_family = AF_INET6;
3070 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3071 rtm->rtm_src_len = rt->rt6i_src.plen;
3074 table = rt->rt6i_table->tb6_id;
3076 table = RT6_TABLE_UNSPEC;
3077 rtm->rtm_table = table;
3078 if (nla_put_u32(skb, RTA_TABLE, table))
3079 goto nla_put_failure;
3080 if (rt->rt6i_flags & RTF_REJECT) {
3081 switch (rt->dst.error) {
3083 rtm->rtm_type = RTN_BLACKHOLE;
3086 rtm->rtm_type = RTN_PROHIBIT;
3089 rtm->rtm_type = RTN_THROW;
3092 rtm->rtm_type = RTN_UNREACHABLE;
3096 else if (rt->rt6i_flags & RTF_LOCAL)
3097 rtm->rtm_type = RTN_LOCAL;
3098 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3099 rtm->rtm_type = RTN_LOCAL;
3101 rtm->rtm_type = RTN_UNICAST;
3103 if (!netif_carrier_ok(rt->dst.dev)) {
3104 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3105 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3106 rtm->rtm_flags |= RTNH_F_DEAD;
3108 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3109 rtm->rtm_protocol = rt->rt6i_protocol;
3110 if (rt->rt6i_flags & RTF_DYNAMIC)
3111 rtm->rtm_protocol = RTPROT_REDIRECT;
3112 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3113 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3114 rtm->rtm_protocol = RTPROT_RA;
3116 rtm->rtm_protocol = RTPROT_KERNEL;
3119 if (rt->rt6i_flags & RTF_CACHE)
3120 rtm->rtm_flags |= RTM_F_CLONED;
3123 if (nla_put_in6_addr(skb, RTA_DST, dst))
3124 goto nla_put_failure;
3125 rtm->rtm_dst_len = 128;
3126 } else if (rtm->rtm_dst_len)
3127 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3128 goto nla_put_failure;
3129 #ifdef CONFIG_IPV6_SUBTREES
3131 if (nla_put_in6_addr(skb, RTA_SRC, src))
3132 goto nla_put_failure;
3133 rtm->rtm_src_len = 128;
3134 } else if (rtm->rtm_src_len &&
3135 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3136 goto nla_put_failure;
3139 #ifdef CONFIG_IPV6_MROUTE
3140 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3141 int err = ip6mr_get_route(net, skb, rtm, nowait);
3146 goto nla_put_failure;
3148 if (err == -EMSGSIZE)
3149 goto nla_put_failure;
3154 if (nla_put_u32(skb, RTA_IIF, iif))
3155 goto nla_put_failure;
3157 struct in6_addr saddr_buf;
3158 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3159 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3160 goto nla_put_failure;
3163 if (rt->rt6i_prefsrc.plen) {
3164 struct in6_addr saddr_buf;
3165 saddr_buf = rt->rt6i_prefsrc.addr;
3166 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3167 goto nla_put_failure;
3170 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3172 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3173 if (rtnetlink_put_metrics(skb, metrics) < 0)
3174 goto nla_put_failure;
3176 if (rt->rt6i_flags & RTF_GATEWAY) {
3177 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3178 goto nla_put_failure;
3182 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3183 goto nla_put_failure;
3184 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3185 goto nla_put_failure;
3187 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3189 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3190 goto nla_put_failure;
3192 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3193 goto nla_put_failure;
3195 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3197 nlmsg_end(skb, nlh);
3201 nlmsg_cancel(skb, nlh);
3205 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3207 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3210 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3211 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3212 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3216 return rt6_fill_node(arg->net,
3217 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3218 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3219 prefix, 0, NLM_F_MULTI);
3222 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3224 struct net *net = sock_net(in_skb->sk);
3225 struct nlattr *tb[RTA_MAX+1];
3226 struct rt6_info *rt;
3227 struct sk_buff *skb;
3230 int err, iif = 0, oif = 0;
3232 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3237 memset(&fl6, 0, sizeof(fl6));
3240 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3243 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3247 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3250 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3254 iif = nla_get_u32(tb[RTA_IIF]);
3257 oif = nla_get_u32(tb[RTA_OIF]);
3260 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3263 struct net_device *dev;
3266 dev = __dev_get_by_index(net, iif);
3272 fl6.flowi6_iif = iif;
3274 if (!ipv6_addr_any(&fl6.saddr))
3275 flags |= RT6_LOOKUP_F_HAS_SADDR;
3277 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3280 fl6.flowi6_oif = oif;
3282 if (netif_index_is_l3_master(net, oif)) {
3283 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3284 FLOWI_FLAG_SKIP_NH_OIF;
3287 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3290 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3297 /* Reserve room for dummy headers, this skb can pass
3298 through good chunk of routing engine.
3300 skb_reset_mac_header(skb);
3301 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3303 skb_dst_set(skb, &rt->dst);
3305 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3306 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3307 nlh->nlmsg_seq, 0, 0, 0);
3313 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3318 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3319 unsigned int nlm_flags)
3321 struct sk_buff *skb;
3322 struct net *net = info->nl_net;
3327 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3329 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3333 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3334 event, info->portid, seq, 0, 0, nlm_flags);
3336 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3337 WARN_ON(err == -EMSGSIZE);
3341 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3342 info->nlh, gfp_any());
3346 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3349 static int ip6_route_dev_notify(struct notifier_block *this,
3350 unsigned long event, void *ptr)
3352 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3353 struct net *net = dev_net(dev);
3355 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3356 net->ipv6.ip6_null_entry->dst.dev = dev;
3357 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3358 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3359 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3360 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3361 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3362 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3373 #ifdef CONFIG_PROC_FS
3375 static const struct file_operations ipv6_route_proc_fops = {
3376 .owner = THIS_MODULE,
3377 .open = ipv6_route_open,
3379 .llseek = seq_lseek,
3380 .release = seq_release_net,
3383 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3385 struct net *net = (struct net *)seq->private;
3386 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3387 net->ipv6.rt6_stats->fib_nodes,
3388 net->ipv6.rt6_stats->fib_route_nodes,
3389 net->ipv6.rt6_stats->fib_rt_alloc,
3390 net->ipv6.rt6_stats->fib_rt_entries,
3391 net->ipv6.rt6_stats->fib_rt_cache,
3392 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3393 net->ipv6.rt6_stats->fib_discarded_routes);
3398 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3400 return single_open_net(inode, file, rt6_stats_seq_show);
3403 static const struct file_operations rt6_stats_seq_fops = {
3404 .owner = THIS_MODULE,
3405 .open = rt6_stats_seq_open,
3407 .llseek = seq_lseek,
3408 .release = single_release_net,
3410 #endif /* CONFIG_PROC_FS */
3412 #ifdef CONFIG_SYSCTL
3415 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3416 void __user *buffer, size_t *lenp, loff_t *ppos)
3423 net = (struct net *)ctl->extra1;
3424 delay = net->ipv6.sysctl.flush_delay;
3425 proc_dointvec(ctl, write, buffer, lenp, ppos);
3426 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3430 struct ctl_table ipv6_route_table_template[] = {
3432 .procname = "flush",
3433 .data = &init_net.ipv6.sysctl.flush_delay,
3434 .maxlen = sizeof(int),
3436 .proc_handler = ipv6_sysctl_rtcache_flush
3439 .procname = "gc_thresh",
3440 .data = &ip6_dst_ops_template.gc_thresh,
3441 .maxlen = sizeof(int),
3443 .proc_handler = proc_dointvec,
3446 .procname = "max_size",
3447 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3448 .maxlen = sizeof(int),
3450 .proc_handler = proc_dointvec,
3453 .procname = "gc_min_interval",
3454 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3455 .maxlen = sizeof(int),
3457 .proc_handler = proc_dointvec_jiffies,
3460 .procname = "gc_timeout",
3461 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3462 .maxlen = sizeof(int),
3464 .proc_handler = proc_dointvec_jiffies,
3467 .procname = "gc_interval",
3468 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3469 .maxlen = sizeof(int),
3471 .proc_handler = proc_dointvec_jiffies,
3474 .procname = "gc_elasticity",
3475 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3476 .maxlen = sizeof(int),
3478 .proc_handler = proc_dointvec,
3481 .procname = "mtu_expires",
3482 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3483 .maxlen = sizeof(int),
3485 .proc_handler = proc_dointvec_jiffies,
3488 .procname = "min_adv_mss",
3489 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3490 .maxlen = sizeof(int),
3492 .proc_handler = proc_dointvec,
3495 .procname = "gc_min_interval_ms",
3496 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3497 .maxlen = sizeof(int),
3499 .proc_handler = proc_dointvec_ms_jiffies,
3504 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3506 struct ctl_table *table;
3508 table = kmemdup(ipv6_route_table_template,
3509 sizeof(ipv6_route_table_template),
3513 table[0].data = &net->ipv6.sysctl.flush_delay;
3514 table[0].extra1 = net;
3515 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3516 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3517 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3518 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3519 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3520 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3521 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3522 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3523 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3525 /* Don't export sysctls to unprivileged users */
3526 if (net->user_ns != &init_user_ns)
3527 table[0].procname = NULL;
3534 static int __net_init ip6_route_net_init(struct net *net)
3538 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3539 sizeof(net->ipv6.ip6_dst_ops));
3541 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3542 goto out_ip6_dst_ops;
3544 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3545 sizeof(*net->ipv6.ip6_null_entry),
3547 if (!net->ipv6.ip6_null_entry)
3548 goto out_ip6_dst_entries;
3549 net->ipv6.ip6_null_entry->dst.path =
3550 (struct dst_entry *)net->ipv6.ip6_null_entry;
3551 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3552 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3553 ip6_template_metrics, true);
3555 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3556 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3557 sizeof(*net->ipv6.ip6_prohibit_entry),
3559 if (!net->ipv6.ip6_prohibit_entry)
3560 goto out_ip6_null_entry;
3561 net->ipv6.ip6_prohibit_entry->dst.path =
3562 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3563 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3564 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3565 ip6_template_metrics, true);
3567 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3568 sizeof(*net->ipv6.ip6_blk_hole_entry),
3570 if (!net->ipv6.ip6_blk_hole_entry)
3571 goto out_ip6_prohibit_entry;
3572 net->ipv6.ip6_blk_hole_entry->dst.path =
3573 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3574 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3575 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3576 ip6_template_metrics, true);
3579 net->ipv6.sysctl.flush_delay = 0;
3580 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3581 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3582 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3583 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3584 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3585 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3586 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3588 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3594 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3595 out_ip6_prohibit_entry:
3596 kfree(net->ipv6.ip6_prohibit_entry);
3598 kfree(net->ipv6.ip6_null_entry);
3600 out_ip6_dst_entries:
3601 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3606 static void __net_exit ip6_route_net_exit(struct net *net)
3608 kfree(net->ipv6.ip6_null_entry);
3609 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3610 kfree(net->ipv6.ip6_prohibit_entry);
3611 kfree(net->ipv6.ip6_blk_hole_entry);
3613 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3616 static int __net_init ip6_route_net_init_late(struct net *net)
3618 #ifdef CONFIG_PROC_FS
3619 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3620 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3625 static void __net_exit ip6_route_net_exit_late(struct net *net)
3627 #ifdef CONFIG_PROC_FS
3628 remove_proc_entry("ipv6_route", net->proc_net);
3629 remove_proc_entry("rt6_stats", net->proc_net);
3633 static struct pernet_operations ip6_route_net_ops = {
3634 .init = ip6_route_net_init,
3635 .exit = ip6_route_net_exit,
3638 static int __net_init ipv6_inetpeer_init(struct net *net)
3640 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3644 inet_peer_base_init(bp);
3645 net->ipv6.peers = bp;
3649 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3651 struct inet_peer_base *bp = net->ipv6.peers;
3653 net->ipv6.peers = NULL;
3654 inetpeer_invalidate_tree(bp);
3658 static struct pernet_operations ipv6_inetpeer_ops = {
3659 .init = ipv6_inetpeer_init,
3660 .exit = ipv6_inetpeer_exit,
3663 static struct pernet_operations ip6_route_net_late_ops = {
3664 .init = ip6_route_net_init_late,
3665 .exit = ip6_route_net_exit_late,
3668 static struct notifier_block ip6_route_dev_notifier = {
3669 .notifier_call = ip6_route_dev_notify,
3673 int __init ip6_route_init(void)
3679 ip6_dst_ops_template.kmem_cachep =
3680 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3681 SLAB_HWCACHE_ALIGN, NULL);
3682 if (!ip6_dst_ops_template.kmem_cachep)
3685 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3687 goto out_kmem_cache;
3689 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3691 goto out_dst_entries;
3693 ret = register_pernet_subsys(&ip6_route_net_ops);
3695 goto out_register_inetpeer;
3697 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3699 /* Registering of the loopback is done before this portion of code,
3700 * the loopback reference in rt6_info will not be taken, do it
3701 * manually for init_net */
3702 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3703 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3704 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3705 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3706 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3707 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3708 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3712 goto out_register_subsys;
3718 ret = fib6_rules_init();
3722 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3724 goto fib6_rules_init;
3727 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3728 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3729 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3730 goto out_register_late_subsys;
3732 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3734 goto out_register_late_subsys;
3736 for_each_possible_cpu(cpu) {
3737 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3739 INIT_LIST_HEAD(&ul->head);
3740 spin_lock_init(&ul->lock);
3746 out_register_late_subsys:
3747 unregister_pernet_subsys(&ip6_route_net_late_ops);
3749 fib6_rules_cleanup();
3754 out_register_subsys:
3755 unregister_pernet_subsys(&ip6_route_net_ops);
3756 out_register_inetpeer:
3757 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3759 dst_entries_destroy(&ip6_dst_blackhole_ops);
3761 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3765 void ip6_route_cleanup(void)
3767 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3768 unregister_pernet_subsys(&ip6_route_net_late_ops);
3769 fib6_rules_cleanup();
3772 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3773 unregister_pernet_subsys(&ip6_route_net_ops);
3774 dst_entries_destroy(&ip6_dst_blackhole_ops);
3775 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);