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_HOPLIMIT && val > 255)
1732 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1736 __set_bit(type - 1, mxc->mx_valid);
1740 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1741 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1751 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1753 struct net *net = cfg->fc_nlinfo.nl_net;
1754 struct rt6_info *rt = NULL;
1755 struct net_device *dev = NULL;
1756 struct inet6_dev *idev = NULL;
1757 struct fib6_table *table;
1761 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1763 #ifndef CONFIG_IPV6_SUBTREES
1764 if (cfg->fc_src_len)
1767 if (cfg->fc_ifindex) {
1769 dev = dev_get_by_index(net, cfg->fc_ifindex);
1772 idev = in6_dev_get(dev);
1777 if (cfg->fc_metric == 0)
1778 cfg->fc_metric = IP6_RT_PRIO_USER;
1781 if (cfg->fc_nlinfo.nlh &&
1782 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1783 table = fib6_get_table(net, cfg->fc_table);
1785 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1786 table = fib6_new_table(net, cfg->fc_table);
1789 table = fib6_new_table(net, cfg->fc_table);
1795 rt = ip6_dst_alloc(net, NULL,
1796 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1803 if (cfg->fc_flags & RTF_EXPIRES)
1804 rt6_set_expires(rt, jiffies +
1805 clock_t_to_jiffies(cfg->fc_expires));
1807 rt6_clean_expires(rt);
1809 if (cfg->fc_protocol == RTPROT_UNSPEC)
1810 cfg->fc_protocol = RTPROT_BOOT;
1811 rt->rt6i_protocol = cfg->fc_protocol;
1813 addr_type = ipv6_addr_type(&cfg->fc_dst);
1815 if (addr_type & IPV6_ADDR_MULTICAST)
1816 rt->dst.input = ip6_mc_input;
1817 else if (cfg->fc_flags & RTF_LOCAL)
1818 rt->dst.input = ip6_input;
1820 rt->dst.input = ip6_forward;
1822 rt->dst.output = ip6_output;
1824 if (cfg->fc_encap) {
1825 struct lwtunnel_state *lwtstate;
1827 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1828 cfg->fc_encap, AF_INET6, cfg,
1832 rt->dst.lwtstate = lwtstate_get(lwtstate);
1833 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1834 rt->dst.lwtstate->orig_output = rt->dst.output;
1835 rt->dst.output = lwtunnel_output;
1837 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1838 rt->dst.lwtstate->orig_input = rt->dst.input;
1839 rt->dst.input = lwtunnel_input;
1843 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1844 rt->rt6i_dst.plen = cfg->fc_dst_len;
1845 if (rt->rt6i_dst.plen == 128)
1846 rt->dst.flags |= DST_HOST;
1848 #ifdef CONFIG_IPV6_SUBTREES
1849 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1850 rt->rt6i_src.plen = cfg->fc_src_len;
1853 rt->rt6i_metric = cfg->fc_metric;
1855 /* We cannot add true routes via loopback here,
1856 they would result in kernel looping; promote them to reject routes
1858 if ((cfg->fc_flags & RTF_REJECT) ||
1859 (dev && (dev->flags & IFF_LOOPBACK) &&
1860 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1861 !(cfg->fc_flags & RTF_LOCAL))) {
1862 /* hold loopback dev/idev if we haven't done so. */
1863 if (dev != net->loopback_dev) {
1868 dev = net->loopback_dev;
1870 idev = in6_dev_get(dev);
1876 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1877 switch (cfg->fc_type) {
1879 rt->dst.error = -EINVAL;
1880 rt->dst.output = dst_discard_out;
1881 rt->dst.input = dst_discard;
1884 rt->dst.error = -EACCES;
1885 rt->dst.output = ip6_pkt_prohibit_out;
1886 rt->dst.input = ip6_pkt_prohibit;
1889 case RTN_UNREACHABLE:
1891 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1892 : (cfg->fc_type == RTN_UNREACHABLE)
1893 ? -EHOSTUNREACH : -ENETUNREACH;
1894 rt->dst.output = ip6_pkt_discard_out;
1895 rt->dst.input = ip6_pkt_discard;
1901 if (cfg->fc_flags & RTF_GATEWAY) {
1902 const struct in6_addr *gw_addr;
1905 gw_addr = &cfg->fc_gateway;
1906 gwa_type = ipv6_addr_type(gw_addr);
1908 /* if gw_addr is local we will fail to detect this in case
1909 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1910 * will return already-added prefix route via interface that
1911 * prefix route was assigned to, which might be non-loopback.
1914 if (ipv6_chk_addr_and_flags(net, gw_addr,
1915 gwa_type & IPV6_ADDR_LINKLOCAL ?
1919 rt->rt6i_gateway = *gw_addr;
1921 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1922 struct rt6_info *grt;
1924 /* IPv6 strictly inhibits using not link-local
1925 addresses as nexthop address.
1926 Otherwise, router will not able to send redirects.
1927 It is very good, but in some (rare!) circumstances
1928 (SIT, PtP, NBMA NOARP links) it is handy to allow
1929 some exceptions. --ANK
1931 if (!(gwa_type & IPV6_ADDR_UNICAST))
1934 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1936 err = -EHOSTUNREACH;
1940 if (dev != grt->dst.dev) {
1946 idev = grt->rt6i_idev;
1948 in6_dev_hold(grt->rt6i_idev);
1950 if (!(grt->rt6i_flags & RTF_GATEWAY))
1958 if (!dev || (dev->flags & IFF_LOOPBACK))
1966 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1967 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1971 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1972 rt->rt6i_prefsrc.plen = 128;
1974 rt->rt6i_prefsrc.plen = 0;
1976 rt->rt6i_flags = cfg->fc_flags;
1980 rt->rt6i_idev = idev;
1981 rt->rt6i_table = table;
1983 cfg->fc_nlinfo.nl_net = dev_net(dev);
1994 return ERR_PTR(err);
1997 int ip6_route_add(struct fib6_config *cfg)
1999 struct mx6_config mxc = { .mx = NULL, };
2000 struct rt6_info *rt;
2003 rt = ip6_route_info_create(cfg);
2010 err = ip6_convert_metrics(&mxc, cfg);
2014 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2026 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2029 struct fib6_table *table;
2030 struct net *net = dev_net(rt->dst.dev);
2032 if (rt == net->ipv6.ip6_null_entry ||
2033 rt->dst.flags & DST_NOCACHE) {
2038 table = rt->rt6i_table;
2039 write_lock_bh(&table->tb6_lock);
2040 err = fib6_del(rt, info);
2041 write_unlock_bh(&table->tb6_lock);
2048 int ip6_del_rt(struct rt6_info *rt)
2050 struct nl_info info = {
2051 .nl_net = dev_net(rt->dst.dev),
2053 return __ip6_del_rt(rt, &info);
2056 static int ip6_route_del(struct fib6_config *cfg)
2058 struct fib6_table *table;
2059 struct fib6_node *fn;
2060 struct rt6_info *rt;
2063 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2067 read_lock_bh(&table->tb6_lock);
2069 fn = fib6_locate(&table->tb6_root,
2070 &cfg->fc_dst, cfg->fc_dst_len,
2071 &cfg->fc_src, cfg->fc_src_len);
2074 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2075 if ((rt->rt6i_flags & RTF_CACHE) &&
2076 !(cfg->fc_flags & RTF_CACHE))
2078 if (cfg->fc_ifindex &&
2080 rt->dst.dev->ifindex != cfg->fc_ifindex))
2082 if (cfg->fc_flags & RTF_GATEWAY &&
2083 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2085 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2088 read_unlock_bh(&table->tb6_lock);
2090 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2093 read_unlock_bh(&table->tb6_lock);
2098 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2100 struct netevent_redirect netevent;
2101 struct rt6_info *rt, *nrt = NULL;
2102 struct ndisc_options ndopts;
2103 struct inet6_dev *in6_dev;
2104 struct neighbour *neigh;
2106 int optlen, on_link;
2109 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2110 optlen -= sizeof(*msg);
2113 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2117 msg = (struct rd_msg *)icmp6_hdr(skb);
2119 if (ipv6_addr_is_multicast(&msg->dest)) {
2120 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2125 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2127 } else if (ipv6_addr_type(&msg->target) !=
2128 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2129 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2133 in6_dev = __in6_dev_get(skb->dev);
2136 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2140 * The IP source address of the Redirect MUST be the same as the current
2141 * first-hop router for the specified ICMP Destination Address.
2144 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2145 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2150 if (ndopts.nd_opts_tgt_lladdr) {
2151 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2154 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2159 rt = (struct rt6_info *) dst;
2160 if (rt->rt6i_flags & RTF_REJECT) {
2161 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2165 /* Redirect received -> path was valid.
2166 * Look, redirects are sent only in response to data packets,
2167 * so that this nexthop apparently is reachable. --ANK
2169 dst_confirm(&rt->dst);
2171 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2176 * We have finally decided to accept it.
2179 neigh_update(neigh, lladdr, NUD_STALE,
2180 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2181 NEIGH_UPDATE_F_OVERRIDE|
2182 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2183 NEIGH_UPDATE_F_ISROUTER))
2186 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2190 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2192 nrt->rt6i_flags &= ~RTF_GATEWAY;
2194 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2196 if (ip6_ins_rt(nrt))
2199 netevent.old = &rt->dst;
2200 netevent.new = &nrt->dst;
2201 netevent.daddr = &msg->dest;
2202 netevent.neigh = neigh;
2203 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2205 if (rt->rt6i_flags & RTF_CACHE) {
2206 rt = (struct rt6_info *) dst_clone(&rt->dst);
2211 neigh_release(neigh);
2215 * Misc support functions
2218 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2220 BUG_ON(from->dst.from);
2222 rt->rt6i_flags &= ~RTF_EXPIRES;
2223 dst_hold(&from->dst);
2224 rt->dst.from = &from->dst;
2225 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2228 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2230 rt->dst.input = ort->dst.input;
2231 rt->dst.output = ort->dst.output;
2232 rt->rt6i_dst = ort->rt6i_dst;
2233 rt->dst.error = ort->dst.error;
2234 rt->rt6i_idev = ort->rt6i_idev;
2236 in6_dev_hold(rt->rt6i_idev);
2237 rt->dst.lastuse = jiffies;
2238 rt->rt6i_gateway = ort->rt6i_gateway;
2239 rt->rt6i_flags = ort->rt6i_flags;
2240 rt6_set_from(rt, ort);
2241 rt->rt6i_metric = ort->rt6i_metric;
2242 #ifdef CONFIG_IPV6_SUBTREES
2243 rt->rt6i_src = ort->rt6i_src;
2245 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2246 rt->rt6i_table = ort->rt6i_table;
2247 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2250 #ifdef CONFIG_IPV6_ROUTE_INFO
2251 static struct rt6_info *rt6_get_route_info(struct net *net,
2252 const struct in6_addr *prefix, int prefixlen,
2253 const struct in6_addr *gwaddr, int ifindex)
2255 struct fib6_node *fn;
2256 struct rt6_info *rt = NULL;
2257 struct fib6_table *table;
2259 table = fib6_get_table(net, RT6_TABLE_INFO);
2263 read_lock_bh(&table->tb6_lock);
2264 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2268 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2269 if (rt->dst.dev->ifindex != ifindex)
2271 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2273 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2279 read_unlock_bh(&table->tb6_lock);
2283 static struct rt6_info *rt6_add_route_info(struct net *net,
2284 const struct in6_addr *prefix, int prefixlen,
2285 const struct in6_addr *gwaddr, int ifindex,
2288 struct fib6_config cfg = {
2289 .fc_metric = IP6_RT_PRIO_USER,
2290 .fc_ifindex = ifindex,
2291 .fc_dst_len = prefixlen,
2292 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2293 RTF_UP | RTF_PREF(pref),
2294 .fc_nlinfo.portid = 0,
2295 .fc_nlinfo.nlh = NULL,
2296 .fc_nlinfo.nl_net = net,
2299 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2300 cfg.fc_dst = *prefix;
2301 cfg.fc_gateway = *gwaddr;
2303 /* We should treat it as a default route if prefix length is 0. */
2305 cfg.fc_flags |= RTF_DEFAULT;
2307 ip6_route_add(&cfg);
2309 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2313 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2315 struct rt6_info *rt;
2316 struct fib6_table *table;
2318 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2322 read_lock_bh(&table->tb6_lock);
2323 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2324 if (dev == rt->dst.dev &&
2325 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2326 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2331 read_unlock_bh(&table->tb6_lock);
2335 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2336 struct net_device *dev,
2339 struct fib6_config cfg = {
2340 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2341 .fc_metric = IP6_RT_PRIO_USER,
2342 .fc_ifindex = dev->ifindex,
2343 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2344 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2345 .fc_nlinfo.portid = 0,
2346 .fc_nlinfo.nlh = NULL,
2347 .fc_nlinfo.nl_net = dev_net(dev),
2350 cfg.fc_gateway = *gwaddr;
2352 ip6_route_add(&cfg);
2354 return rt6_get_dflt_router(gwaddr, dev);
2357 void rt6_purge_dflt_routers(struct net *net)
2359 struct rt6_info *rt;
2360 struct fib6_table *table;
2362 /* NOTE: Keep consistent with rt6_get_dflt_router */
2363 table = fib6_get_table(net, RT6_TABLE_DFLT);
2368 read_lock_bh(&table->tb6_lock);
2369 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2370 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2371 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2373 read_unlock_bh(&table->tb6_lock);
2378 read_unlock_bh(&table->tb6_lock);
2381 static void rtmsg_to_fib6_config(struct net *net,
2382 struct in6_rtmsg *rtmsg,
2383 struct fib6_config *cfg)
2385 memset(cfg, 0, sizeof(*cfg));
2387 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2389 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2390 cfg->fc_metric = rtmsg->rtmsg_metric;
2391 cfg->fc_expires = rtmsg->rtmsg_info;
2392 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2393 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2394 cfg->fc_flags = rtmsg->rtmsg_flags;
2396 cfg->fc_nlinfo.nl_net = net;
2398 cfg->fc_dst = rtmsg->rtmsg_dst;
2399 cfg->fc_src = rtmsg->rtmsg_src;
2400 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2403 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2405 struct fib6_config cfg;
2406 struct in6_rtmsg rtmsg;
2410 case SIOCADDRT: /* Add a route */
2411 case SIOCDELRT: /* Delete a route */
2412 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2414 err = copy_from_user(&rtmsg, arg,
2415 sizeof(struct in6_rtmsg));
2419 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2424 err = ip6_route_add(&cfg);
2427 err = ip6_route_del(&cfg);
2441 * Drop the packet on the floor
2444 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2447 struct dst_entry *dst = skb_dst(skb);
2448 switch (ipstats_mib_noroutes) {
2449 case IPSTATS_MIB_INNOROUTES:
2450 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2451 if (type == IPV6_ADDR_ANY) {
2452 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2453 IPSTATS_MIB_INADDRERRORS);
2457 case IPSTATS_MIB_OUTNOROUTES:
2458 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2459 ipstats_mib_noroutes);
2462 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2467 static int ip6_pkt_discard(struct sk_buff *skb)
2469 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2472 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2474 skb->dev = skb_dst(skb)->dev;
2475 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2478 static int ip6_pkt_prohibit(struct sk_buff *skb)
2480 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2483 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2485 skb->dev = skb_dst(skb)->dev;
2486 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2490 * Allocate a dst for local (unicast / anycast) address.
2493 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2494 const struct in6_addr *addr,
2498 struct net *net = dev_net(idev->dev);
2499 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2502 return ERR_PTR(-ENOMEM);
2506 rt->dst.flags |= DST_HOST;
2507 rt->dst.input = ip6_input;
2508 rt->dst.output = ip6_output;
2509 rt->rt6i_idev = idev;
2511 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2513 rt->rt6i_flags |= RTF_ANYCAST;
2515 rt->rt6i_flags |= RTF_LOCAL;
2517 rt->rt6i_gateway = *addr;
2518 rt->rt6i_dst.addr = *addr;
2519 rt->rt6i_dst.plen = 128;
2520 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2521 rt->rt6i_table = fib6_get_table(net, tb_id);
2522 rt->dst.flags |= DST_NOCACHE;
2524 atomic_set(&rt->dst.__refcnt, 1);
2529 int ip6_route_get_saddr(struct net *net,
2530 struct rt6_info *rt,
2531 const struct in6_addr *daddr,
2533 struct in6_addr *saddr)
2535 struct inet6_dev *idev =
2536 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2538 if (rt && rt->rt6i_prefsrc.plen)
2539 *saddr = rt->rt6i_prefsrc.addr;
2541 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2542 daddr, prefs, saddr);
2546 /* remove deleted ip from prefsrc entries */
2547 struct arg_dev_net_ip {
2548 struct net_device *dev;
2550 struct in6_addr *addr;
2553 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2555 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2556 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2557 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2559 if (((void *)rt->dst.dev == dev || !dev) &&
2560 rt != net->ipv6.ip6_null_entry &&
2561 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2562 /* remove prefsrc entry */
2563 rt->rt6i_prefsrc.plen = 0;
2568 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2570 struct net *net = dev_net(ifp->idev->dev);
2571 struct arg_dev_net_ip adni = {
2572 .dev = ifp->idev->dev,
2576 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2579 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2580 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2582 /* Remove routers and update dst entries when gateway turn into host. */
2583 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2585 struct in6_addr *gateway = (struct in6_addr *)arg;
2587 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2588 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2589 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2595 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2597 fib6_clean_all(net, fib6_clean_tohost, gateway);
2600 struct arg_dev_net {
2601 struct net_device *dev;
2605 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2607 const struct arg_dev_net *adn = arg;
2608 const struct net_device *dev = adn->dev;
2610 if ((rt->dst.dev == dev || !dev) &&
2611 rt != adn->net->ipv6.ip6_null_entry)
2617 void rt6_ifdown(struct net *net, struct net_device *dev)
2619 struct arg_dev_net adn = {
2624 fib6_clean_all(net, fib6_ifdown, &adn);
2625 icmp6_clean_all(fib6_ifdown, &adn);
2627 rt6_uncached_list_flush_dev(net, dev);
2630 struct rt6_mtu_change_arg {
2631 struct net_device *dev;
2635 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2637 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2638 struct inet6_dev *idev;
2640 /* In IPv6 pmtu discovery is not optional,
2641 so that RTAX_MTU lock cannot disable it.
2642 We still use this lock to block changes
2643 caused by addrconf/ndisc.
2646 idev = __in6_dev_get(arg->dev);
2650 /* For administrative MTU increase, there is no way to discover
2651 IPv6 PMTU increase, so PMTU increase should be updated here.
2652 Since RFC 1981 doesn't include administrative MTU increase
2653 update PMTU increase is a MUST. (i.e. jumbo frame)
2656 If new MTU is less than route PMTU, this new MTU will be the
2657 lowest MTU in the path, update the route PMTU to reflect PMTU
2658 decreases; if new MTU is greater than route PMTU, and the
2659 old MTU is the lowest MTU in the path, update the route PMTU
2660 to reflect the increase. In this case if the other nodes' MTU
2661 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2664 if (rt->dst.dev == arg->dev &&
2665 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2666 if (rt->rt6i_flags & RTF_CACHE) {
2667 /* For RTF_CACHE with rt6i_pmtu == 0
2668 * (i.e. a redirected route),
2669 * the metrics of its rt->dst.from has already
2672 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2673 rt->rt6i_pmtu = arg->mtu;
2674 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2675 (dst_mtu(&rt->dst) < arg->mtu &&
2676 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2677 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2683 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2685 struct rt6_mtu_change_arg arg = {
2690 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2693 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2694 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2695 [RTA_OIF] = { .type = NLA_U32 },
2696 [RTA_IIF] = { .type = NLA_U32 },
2697 [RTA_PRIORITY] = { .type = NLA_U32 },
2698 [RTA_METRICS] = { .type = NLA_NESTED },
2699 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2700 [RTA_PREF] = { .type = NLA_U8 },
2701 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2702 [RTA_ENCAP] = { .type = NLA_NESTED },
2705 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2706 struct fib6_config *cfg)
2709 struct nlattr *tb[RTA_MAX+1];
2713 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2718 rtm = nlmsg_data(nlh);
2719 memset(cfg, 0, sizeof(*cfg));
2721 cfg->fc_table = rtm->rtm_table;
2722 cfg->fc_dst_len = rtm->rtm_dst_len;
2723 cfg->fc_src_len = rtm->rtm_src_len;
2724 cfg->fc_flags = RTF_UP;
2725 cfg->fc_protocol = rtm->rtm_protocol;
2726 cfg->fc_type = rtm->rtm_type;
2728 if (rtm->rtm_type == RTN_UNREACHABLE ||
2729 rtm->rtm_type == RTN_BLACKHOLE ||
2730 rtm->rtm_type == RTN_PROHIBIT ||
2731 rtm->rtm_type == RTN_THROW)
2732 cfg->fc_flags |= RTF_REJECT;
2734 if (rtm->rtm_type == RTN_LOCAL)
2735 cfg->fc_flags |= RTF_LOCAL;
2737 if (rtm->rtm_flags & RTM_F_CLONED)
2738 cfg->fc_flags |= RTF_CACHE;
2740 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2741 cfg->fc_nlinfo.nlh = nlh;
2742 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2744 if (tb[RTA_GATEWAY]) {
2745 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2746 cfg->fc_flags |= RTF_GATEWAY;
2750 int plen = (rtm->rtm_dst_len + 7) >> 3;
2752 if (nla_len(tb[RTA_DST]) < plen)
2755 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2759 int plen = (rtm->rtm_src_len + 7) >> 3;
2761 if (nla_len(tb[RTA_SRC]) < plen)
2764 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2767 if (tb[RTA_PREFSRC])
2768 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2771 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2773 if (tb[RTA_PRIORITY])
2774 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2776 if (tb[RTA_METRICS]) {
2777 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2778 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2782 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2784 if (tb[RTA_MULTIPATH]) {
2785 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2786 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2790 pref = nla_get_u8(tb[RTA_PREF]);
2791 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2792 pref != ICMPV6_ROUTER_PREF_HIGH)
2793 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2794 cfg->fc_flags |= RTF_PREF(pref);
2798 cfg->fc_encap = tb[RTA_ENCAP];
2800 if (tb[RTA_ENCAP_TYPE])
2801 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2809 struct rt6_info *rt6_info;
2810 struct fib6_config r_cfg;
2811 struct mx6_config mxc;
2812 struct list_head next;
2815 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2819 list_for_each_entry(nh, rt6_nh_list, next) {
2820 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2821 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2822 nh->r_cfg.fc_ifindex);
2826 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2827 struct rt6_info *rt, struct fib6_config *r_cfg)
2830 struct rt6_info *rtnh;
2833 list_for_each_entry(nh, rt6_nh_list, next) {
2834 /* check if rt6_info already exists */
2835 rtnh = nh->rt6_info;
2837 if (rtnh->dst.dev == rt->dst.dev &&
2838 rtnh->rt6i_idev == rt->rt6i_idev &&
2839 ipv6_addr_equal(&rtnh->rt6i_gateway,
2844 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2848 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2853 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2854 list_add_tail(&nh->next, rt6_nh_list);
2859 static int ip6_route_multipath_add(struct fib6_config *cfg)
2861 struct fib6_config r_cfg;
2862 struct rtnexthop *rtnh;
2863 struct rt6_info *rt;
2864 struct rt6_nh *err_nh;
2865 struct rt6_nh *nh, *nh_safe;
2870 int replace = (cfg->fc_nlinfo.nlh &&
2871 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2872 LIST_HEAD(rt6_nh_list);
2874 remaining = cfg->fc_mp_len;
2875 rtnh = (struct rtnexthop *)cfg->fc_mp;
2877 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2878 * rt6_info structs per nexthop
2880 while (rtnh_ok(rtnh, remaining)) {
2881 memcpy(&r_cfg, cfg, sizeof(*cfg));
2882 if (rtnh->rtnh_ifindex)
2883 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2885 attrlen = rtnh_attrlen(rtnh);
2887 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2889 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2891 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2892 r_cfg.fc_flags |= RTF_GATEWAY;
2894 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2895 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2897 r_cfg.fc_encap_type = nla_get_u16(nla);
2900 rt = ip6_route_info_create(&r_cfg);
2907 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2913 rtnh = rtnh_next(rtnh, &remaining);
2917 list_for_each_entry(nh, &rt6_nh_list, next) {
2918 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2919 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2920 nh->rt6_info = NULL;
2923 ip6_print_replace_route_err(&rt6_nh_list);
2928 /* Because each route is added like a single route we remove
2929 * these flags after the first nexthop: if there is a collision,
2930 * we have already failed to add the first nexthop:
2931 * fib6_add_rt2node() has rejected it; when replacing, old
2932 * nexthops have been replaced by first new, the rest should
2935 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2943 /* Delete routes that were already added */
2944 list_for_each_entry(nh, &rt6_nh_list, next) {
2947 ip6_route_del(&nh->r_cfg);
2951 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2953 dst_free(&nh->rt6_info->dst);
2955 list_del(&nh->next);
2962 static int ip6_route_multipath_del(struct fib6_config *cfg)
2964 struct fib6_config r_cfg;
2965 struct rtnexthop *rtnh;
2968 int err = 1, last_err = 0;
2970 remaining = cfg->fc_mp_len;
2971 rtnh = (struct rtnexthop *)cfg->fc_mp;
2973 /* Parse a Multipath Entry */
2974 while (rtnh_ok(rtnh, remaining)) {
2975 memcpy(&r_cfg, cfg, sizeof(*cfg));
2976 if (rtnh->rtnh_ifindex)
2977 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2979 attrlen = rtnh_attrlen(rtnh);
2981 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2983 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2985 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2986 r_cfg.fc_flags |= RTF_GATEWAY;
2989 err = ip6_route_del(&r_cfg);
2993 rtnh = rtnh_next(rtnh, &remaining);
2999 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3001 struct fib6_config cfg;
3004 err = rtm_to_fib6_config(skb, nlh, &cfg);
3009 return ip6_route_multipath_del(&cfg);
3011 return ip6_route_del(&cfg);
3014 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3016 struct fib6_config cfg;
3019 err = rtm_to_fib6_config(skb, nlh, &cfg);
3024 return ip6_route_multipath_add(&cfg);
3026 return ip6_route_add(&cfg);
3029 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3031 return NLMSG_ALIGN(sizeof(struct rtmsg))
3032 + nla_total_size(16) /* RTA_SRC */
3033 + nla_total_size(16) /* RTA_DST */
3034 + nla_total_size(16) /* RTA_GATEWAY */
3035 + nla_total_size(16) /* RTA_PREFSRC */
3036 + nla_total_size(4) /* RTA_TABLE */
3037 + nla_total_size(4) /* RTA_IIF */
3038 + nla_total_size(4) /* RTA_OIF */
3039 + nla_total_size(4) /* RTA_PRIORITY */
3040 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3041 + nla_total_size(sizeof(struct rta_cacheinfo))
3042 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3043 + nla_total_size(1) /* RTA_PREF */
3044 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3047 static int rt6_fill_node(struct net *net,
3048 struct sk_buff *skb, struct rt6_info *rt,
3049 struct in6_addr *dst, struct in6_addr *src,
3050 int iif, int type, u32 portid, u32 seq,
3051 int prefix, int nowait, unsigned int flags)
3053 u32 metrics[RTAX_MAX];
3055 struct nlmsghdr *nlh;
3059 if (prefix) { /* user wants prefix routes only */
3060 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3061 /* success since this is not a prefix route */
3066 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3070 rtm = nlmsg_data(nlh);
3071 rtm->rtm_family = AF_INET6;
3072 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3073 rtm->rtm_src_len = rt->rt6i_src.plen;
3076 table = rt->rt6i_table->tb6_id;
3078 table = RT6_TABLE_UNSPEC;
3079 rtm->rtm_table = table;
3080 if (nla_put_u32(skb, RTA_TABLE, table))
3081 goto nla_put_failure;
3082 if (rt->rt6i_flags & RTF_REJECT) {
3083 switch (rt->dst.error) {
3085 rtm->rtm_type = RTN_BLACKHOLE;
3088 rtm->rtm_type = RTN_PROHIBIT;
3091 rtm->rtm_type = RTN_THROW;
3094 rtm->rtm_type = RTN_UNREACHABLE;
3098 else if (rt->rt6i_flags & RTF_LOCAL)
3099 rtm->rtm_type = RTN_LOCAL;
3100 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3101 rtm->rtm_type = RTN_LOCAL;
3103 rtm->rtm_type = RTN_UNICAST;
3105 if (!netif_carrier_ok(rt->dst.dev)) {
3106 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3107 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3108 rtm->rtm_flags |= RTNH_F_DEAD;
3110 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3111 rtm->rtm_protocol = rt->rt6i_protocol;
3112 if (rt->rt6i_flags & RTF_DYNAMIC)
3113 rtm->rtm_protocol = RTPROT_REDIRECT;
3114 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3115 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3116 rtm->rtm_protocol = RTPROT_RA;
3118 rtm->rtm_protocol = RTPROT_KERNEL;
3121 if (rt->rt6i_flags & RTF_CACHE)
3122 rtm->rtm_flags |= RTM_F_CLONED;
3125 if (nla_put_in6_addr(skb, RTA_DST, dst))
3126 goto nla_put_failure;
3127 rtm->rtm_dst_len = 128;
3128 } else if (rtm->rtm_dst_len)
3129 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3130 goto nla_put_failure;
3131 #ifdef CONFIG_IPV6_SUBTREES
3133 if (nla_put_in6_addr(skb, RTA_SRC, src))
3134 goto nla_put_failure;
3135 rtm->rtm_src_len = 128;
3136 } else if (rtm->rtm_src_len &&
3137 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3138 goto nla_put_failure;
3141 #ifdef CONFIG_IPV6_MROUTE
3142 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3143 int err = ip6mr_get_route(net, skb, rtm, nowait,
3150 goto nla_put_failure;
3152 if (err == -EMSGSIZE)
3153 goto nla_put_failure;
3158 if (nla_put_u32(skb, RTA_IIF, iif))
3159 goto nla_put_failure;
3161 struct in6_addr saddr_buf;
3162 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3163 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3164 goto nla_put_failure;
3167 if (rt->rt6i_prefsrc.plen) {
3168 struct in6_addr saddr_buf;
3169 saddr_buf = rt->rt6i_prefsrc.addr;
3170 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3171 goto nla_put_failure;
3174 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3176 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3177 if (rtnetlink_put_metrics(skb, metrics) < 0)
3178 goto nla_put_failure;
3180 if (rt->rt6i_flags & RTF_GATEWAY) {
3181 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3182 goto nla_put_failure;
3186 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3187 goto nla_put_failure;
3188 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3189 goto nla_put_failure;
3191 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3193 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3194 goto nla_put_failure;
3196 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3197 goto nla_put_failure;
3199 if (lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3200 goto nla_put_failure;
3202 nlmsg_end(skb, nlh);
3206 nlmsg_cancel(skb, nlh);
3210 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3212 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3215 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3216 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3217 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3221 return rt6_fill_node(arg->net,
3222 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3223 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3224 prefix, 0, NLM_F_MULTI);
3227 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3229 struct net *net = sock_net(in_skb->sk);
3230 struct nlattr *tb[RTA_MAX+1];
3231 struct rt6_info *rt;
3232 struct sk_buff *skb;
3235 int err, iif = 0, oif = 0;
3237 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3242 memset(&fl6, 0, sizeof(fl6));
3245 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3248 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3252 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3255 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3259 iif = nla_get_u32(tb[RTA_IIF]);
3262 oif = nla_get_u32(tb[RTA_OIF]);
3265 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3268 struct net_device *dev;
3271 dev = __dev_get_by_index(net, iif);
3277 fl6.flowi6_iif = iif;
3279 if (!ipv6_addr_any(&fl6.saddr))
3280 flags |= RT6_LOOKUP_F_HAS_SADDR;
3282 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3285 fl6.flowi6_oif = oif;
3287 if (netif_index_is_l3_master(net, oif)) {
3288 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3289 FLOWI_FLAG_SKIP_NH_OIF;
3292 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3295 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3302 /* Reserve room for dummy headers, this skb can pass
3303 through good chunk of routing engine.
3305 skb_reset_mac_header(skb);
3306 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3308 skb_dst_set(skb, &rt->dst);
3310 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3311 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3312 nlh->nlmsg_seq, 0, 0, 0);
3318 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3323 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3324 unsigned int nlm_flags)
3326 struct sk_buff *skb;
3327 struct net *net = info->nl_net;
3332 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3334 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3338 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3339 event, info->portid, seq, 0, 0, nlm_flags);
3341 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3342 WARN_ON(err == -EMSGSIZE);
3346 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3347 info->nlh, gfp_any());
3351 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3354 static int ip6_route_dev_notify(struct notifier_block *this,
3355 unsigned long event, void *ptr)
3357 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3358 struct net *net = dev_net(dev);
3360 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3361 net->ipv6.ip6_null_entry->dst.dev = dev;
3362 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3363 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3364 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3365 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3366 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3367 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3378 #ifdef CONFIG_PROC_FS
3380 static const struct file_operations ipv6_route_proc_fops = {
3381 .owner = THIS_MODULE,
3382 .open = ipv6_route_open,
3384 .llseek = seq_lseek,
3385 .release = seq_release_net,
3388 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3390 struct net *net = (struct net *)seq->private;
3391 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3392 net->ipv6.rt6_stats->fib_nodes,
3393 net->ipv6.rt6_stats->fib_route_nodes,
3394 net->ipv6.rt6_stats->fib_rt_alloc,
3395 net->ipv6.rt6_stats->fib_rt_entries,
3396 net->ipv6.rt6_stats->fib_rt_cache,
3397 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3398 net->ipv6.rt6_stats->fib_discarded_routes);
3403 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3405 return single_open_net(inode, file, rt6_stats_seq_show);
3408 static const struct file_operations rt6_stats_seq_fops = {
3409 .owner = THIS_MODULE,
3410 .open = rt6_stats_seq_open,
3412 .llseek = seq_lseek,
3413 .release = single_release_net,
3415 #endif /* CONFIG_PROC_FS */
3417 #ifdef CONFIG_SYSCTL
3420 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3421 void __user *buffer, size_t *lenp, loff_t *ppos)
3428 net = (struct net *)ctl->extra1;
3429 delay = net->ipv6.sysctl.flush_delay;
3430 proc_dointvec(ctl, write, buffer, lenp, ppos);
3431 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3435 struct ctl_table ipv6_route_table_template[] = {
3437 .procname = "flush",
3438 .data = &init_net.ipv6.sysctl.flush_delay,
3439 .maxlen = sizeof(int),
3441 .proc_handler = ipv6_sysctl_rtcache_flush
3444 .procname = "gc_thresh",
3445 .data = &ip6_dst_ops_template.gc_thresh,
3446 .maxlen = sizeof(int),
3448 .proc_handler = proc_dointvec,
3451 .procname = "max_size",
3452 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3453 .maxlen = sizeof(int),
3455 .proc_handler = proc_dointvec,
3458 .procname = "gc_min_interval",
3459 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3460 .maxlen = sizeof(int),
3462 .proc_handler = proc_dointvec_jiffies,
3465 .procname = "gc_timeout",
3466 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3467 .maxlen = sizeof(int),
3469 .proc_handler = proc_dointvec_jiffies,
3472 .procname = "gc_interval",
3473 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3474 .maxlen = sizeof(int),
3476 .proc_handler = proc_dointvec_jiffies,
3479 .procname = "gc_elasticity",
3480 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3481 .maxlen = sizeof(int),
3483 .proc_handler = proc_dointvec,
3486 .procname = "mtu_expires",
3487 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3488 .maxlen = sizeof(int),
3490 .proc_handler = proc_dointvec_jiffies,
3493 .procname = "min_adv_mss",
3494 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3495 .maxlen = sizeof(int),
3497 .proc_handler = proc_dointvec,
3500 .procname = "gc_min_interval_ms",
3501 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3502 .maxlen = sizeof(int),
3504 .proc_handler = proc_dointvec_ms_jiffies,
3509 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3511 struct ctl_table *table;
3513 table = kmemdup(ipv6_route_table_template,
3514 sizeof(ipv6_route_table_template),
3518 table[0].data = &net->ipv6.sysctl.flush_delay;
3519 table[0].extra1 = net;
3520 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3521 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3522 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3523 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3524 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3525 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3526 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3527 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3528 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3530 /* Don't export sysctls to unprivileged users */
3531 if (net->user_ns != &init_user_ns)
3532 table[0].procname = NULL;
3539 static int __net_init ip6_route_net_init(struct net *net)
3543 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3544 sizeof(net->ipv6.ip6_dst_ops));
3546 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3547 goto out_ip6_dst_ops;
3549 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3550 sizeof(*net->ipv6.ip6_null_entry),
3552 if (!net->ipv6.ip6_null_entry)
3553 goto out_ip6_dst_entries;
3554 net->ipv6.ip6_null_entry->dst.path =
3555 (struct dst_entry *)net->ipv6.ip6_null_entry;
3556 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3557 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3558 ip6_template_metrics, true);
3560 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3561 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3562 sizeof(*net->ipv6.ip6_prohibit_entry),
3564 if (!net->ipv6.ip6_prohibit_entry)
3565 goto out_ip6_null_entry;
3566 net->ipv6.ip6_prohibit_entry->dst.path =
3567 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3568 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3569 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3570 ip6_template_metrics, true);
3572 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3573 sizeof(*net->ipv6.ip6_blk_hole_entry),
3575 if (!net->ipv6.ip6_blk_hole_entry)
3576 goto out_ip6_prohibit_entry;
3577 net->ipv6.ip6_blk_hole_entry->dst.path =
3578 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3579 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3580 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3581 ip6_template_metrics, true);
3584 net->ipv6.sysctl.flush_delay = 0;
3585 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3586 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3587 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3588 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3589 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3590 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3591 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3593 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3599 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3600 out_ip6_prohibit_entry:
3601 kfree(net->ipv6.ip6_prohibit_entry);
3603 kfree(net->ipv6.ip6_null_entry);
3605 out_ip6_dst_entries:
3606 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3611 static void __net_exit ip6_route_net_exit(struct net *net)
3613 kfree(net->ipv6.ip6_null_entry);
3614 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3615 kfree(net->ipv6.ip6_prohibit_entry);
3616 kfree(net->ipv6.ip6_blk_hole_entry);
3618 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3621 static int __net_init ip6_route_net_init_late(struct net *net)
3623 #ifdef CONFIG_PROC_FS
3624 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3625 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3630 static void __net_exit ip6_route_net_exit_late(struct net *net)
3632 #ifdef CONFIG_PROC_FS
3633 remove_proc_entry("ipv6_route", net->proc_net);
3634 remove_proc_entry("rt6_stats", net->proc_net);
3638 static struct pernet_operations ip6_route_net_ops = {
3639 .init = ip6_route_net_init,
3640 .exit = ip6_route_net_exit,
3643 static int __net_init ipv6_inetpeer_init(struct net *net)
3645 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3649 inet_peer_base_init(bp);
3650 net->ipv6.peers = bp;
3654 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3656 struct inet_peer_base *bp = net->ipv6.peers;
3658 net->ipv6.peers = NULL;
3659 inetpeer_invalidate_tree(bp);
3663 static struct pernet_operations ipv6_inetpeer_ops = {
3664 .init = ipv6_inetpeer_init,
3665 .exit = ipv6_inetpeer_exit,
3668 static struct pernet_operations ip6_route_net_late_ops = {
3669 .init = ip6_route_net_init_late,
3670 .exit = ip6_route_net_exit_late,
3673 static struct notifier_block ip6_route_dev_notifier = {
3674 .notifier_call = ip6_route_dev_notify,
3678 int __init ip6_route_init(void)
3684 ip6_dst_ops_template.kmem_cachep =
3685 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3686 SLAB_HWCACHE_ALIGN, NULL);
3687 if (!ip6_dst_ops_template.kmem_cachep)
3690 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3692 goto out_kmem_cache;
3694 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3696 goto out_dst_entries;
3698 ret = register_pernet_subsys(&ip6_route_net_ops);
3700 goto out_register_inetpeer;
3702 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3704 /* Registering of the loopback is done before this portion of code,
3705 * the loopback reference in rt6_info will not be taken, do it
3706 * manually for init_net */
3707 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3708 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3709 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3710 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3711 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3712 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3713 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3717 goto out_register_subsys;
3723 ret = fib6_rules_init();
3727 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3729 goto fib6_rules_init;
3732 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3733 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3734 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3735 goto out_register_late_subsys;
3737 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3739 goto out_register_late_subsys;
3741 for_each_possible_cpu(cpu) {
3742 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3744 INIT_LIST_HEAD(&ul->head);
3745 spin_lock_init(&ul->lock);
3751 out_register_late_subsys:
3752 unregister_pernet_subsys(&ip6_route_net_late_ops);
3754 fib6_rules_cleanup();
3759 out_register_subsys:
3760 unregister_pernet_subsys(&ip6_route_net_ops);
3761 out_register_inetpeer:
3762 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3764 dst_entries_destroy(&ip6_dst_blackhole_ops);
3766 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3770 void ip6_route_cleanup(void)
3772 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3773 unregister_pernet_subsys(&ip6_route_net_late_ops);
3774 fib6_rules_cleanup();
3777 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3778 unregister_pernet_subsys(&ip6_route_net_ops);
3779 dst_entries_destroy(&ip6_dst_blackhole_ops);
3780 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);