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 for_each_possible_cpu(cpu) {
147 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
150 spin_lock_bh(&ul->lock);
151 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
152 struct inet6_dev *rt_idev = rt->rt6i_idev;
153 struct net_device *rt_dev = rt->dst.dev;
155 if (rt_idev && (rt_idev->dev == dev || !dev) &&
156 rt_idev->dev != loopback_dev) {
157 rt->rt6i_idev = in6_dev_get(loopback_dev);
158 in6_dev_put(rt_idev);
161 if (rt_dev && (rt_dev == dev || !dev) &&
162 rt_dev != loopback_dev) {
163 rt->dst.dev = loopback_dev;
164 dev_hold(rt->dst.dev);
168 spin_unlock_bh(&ul->lock);
172 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
174 return dst_metrics_write_ptr(rt->dst.from);
177 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
179 struct rt6_info *rt = (struct rt6_info *)dst;
181 if (rt->rt6i_flags & RTF_PCPU)
182 return rt6_pcpu_cow_metrics(rt);
183 else if (rt->rt6i_flags & RTF_CACHE)
186 return dst_cow_metrics_generic(dst, old);
189 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
193 struct in6_addr *p = &rt->rt6i_gateway;
195 if (!ipv6_addr_any(p))
196 return (const void *) p;
198 return &ipv6_hdr(skb)->daddr;
202 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
206 struct rt6_info *rt = (struct rt6_info *) dst;
209 daddr = choose_neigh_daddr(rt, skb, daddr);
210 n = __ipv6_neigh_lookup(dst->dev, daddr);
213 return neigh_create(&nd_tbl, daddr, dst->dev);
216 static struct dst_ops ip6_dst_ops_template = {
220 .check = ip6_dst_check,
221 .default_advmss = ip6_default_advmss,
223 .cow_metrics = ipv6_cow_metrics,
224 .destroy = ip6_dst_destroy,
225 .ifdown = ip6_dst_ifdown,
226 .negative_advice = ip6_negative_advice,
227 .link_failure = ip6_link_failure,
228 .update_pmtu = ip6_rt_update_pmtu,
229 .redirect = rt6_do_redirect,
230 .local_out = __ip6_local_out,
231 .neigh_lookup = ip6_neigh_lookup,
234 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
236 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
238 return mtu ? : dst->dev->mtu;
241 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
242 struct sk_buff *skb, u32 mtu)
246 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
251 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
257 static struct dst_ops ip6_dst_blackhole_ops = {
259 .destroy = ip6_dst_destroy,
260 .check = ip6_dst_check,
261 .mtu = ip6_blackhole_mtu,
262 .default_advmss = ip6_default_advmss,
263 .update_pmtu = ip6_rt_blackhole_update_pmtu,
264 .redirect = ip6_rt_blackhole_redirect,
265 .cow_metrics = ip6_rt_blackhole_cow_metrics,
266 .neigh_lookup = ip6_neigh_lookup,
269 static const u32 ip6_template_metrics[RTAX_MAX] = {
270 [RTAX_HOPLIMIT - 1] = 0,
273 static const struct rt6_info ip6_null_entry_template = {
275 .__refcnt = ATOMIC_INIT(1),
277 .obsolete = DST_OBSOLETE_FORCE_CHK,
278 .error = -ENETUNREACH,
279 .input = ip6_pkt_discard,
280 .output = ip6_pkt_discard_out,
282 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
283 .rt6i_protocol = RTPROT_KERNEL,
284 .rt6i_metric = ~(u32) 0,
285 .rt6i_ref = ATOMIC_INIT(1),
288 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
290 static const struct rt6_info ip6_prohibit_entry_template = {
292 .__refcnt = ATOMIC_INIT(1),
294 .obsolete = DST_OBSOLETE_FORCE_CHK,
296 .input = ip6_pkt_prohibit,
297 .output = ip6_pkt_prohibit_out,
299 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
300 .rt6i_protocol = RTPROT_KERNEL,
301 .rt6i_metric = ~(u32) 0,
302 .rt6i_ref = ATOMIC_INIT(1),
305 static const struct rt6_info ip6_blk_hole_entry_template = {
307 .__refcnt = ATOMIC_INIT(1),
309 .obsolete = DST_OBSOLETE_FORCE_CHK,
311 .input = dst_discard,
312 .output = dst_discard_out,
314 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
315 .rt6i_protocol = RTPROT_KERNEL,
316 .rt6i_metric = ~(u32) 0,
317 .rt6i_ref = ATOMIC_INIT(1),
322 /* allocate dst with ip6_dst_ops */
323 static struct rt6_info *__ip6_dst_alloc(struct net *net,
324 struct net_device *dev,
327 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
328 0, DST_OBSOLETE_FORCE_CHK, flags);
331 struct dst_entry *dst = &rt->dst;
333 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
334 INIT_LIST_HEAD(&rt->rt6i_siblings);
335 INIT_LIST_HEAD(&rt->rt6i_uncached);
340 static struct rt6_info *ip6_dst_alloc(struct net *net,
341 struct net_device *dev,
344 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
347 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
351 for_each_possible_cpu(cpu) {
354 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
355 /* no one shares rt */
359 dst_destroy((struct dst_entry *)rt);
367 static void ip6_dst_destroy(struct dst_entry *dst)
369 struct rt6_info *rt = (struct rt6_info *)dst;
370 struct dst_entry *from = dst->from;
371 struct inet6_dev *idev;
373 dst_destroy_metrics_generic(dst);
374 free_percpu(rt->rt6i_pcpu);
375 rt6_uncached_list_del(rt);
377 idev = rt->rt6i_idev;
379 rt->rt6i_idev = NULL;
387 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
390 struct rt6_info *rt = (struct rt6_info *)dst;
391 struct inet6_dev *idev = rt->rt6i_idev;
392 struct net_device *loopback_dev =
393 dev_net(dev)->loopback_dev;
395 if (dev != loopback_dev) {
396 if (idev && idev->dev == dev) {
397 struct inet6_dev *loopback_idev =
398 in6_dev_get(loopback_dev);
400 rt->rt6i_idev = loopback_idev;
407 static bool rt6_check_expired(const struct rt6_info *rt)
409 if (rt->rt6i_flags & RTF_EXPIRES) {
410 if (time_after(jiffies, rt->dst.expires))
412 } else if (rt->dst.from) {
413 return rt6_check_expired((struct rt6_info *) rt->dst.from);
418 /* Multipath route selection:
419 * Hash based function using packet header and flowlabel.
420 * Adapted from fib_info_hashfn()
422 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
423 const struct flowi6 *fl6)
425 return get_hash_from_flowi6(fl6) % candidate_count;
428 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
429 struct flowi6 *fl6, int oif,
432 struct rt6_info *sibling, *next_sibling;
435 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
436 /* Don't change the route, if route_choosen == 0
437 * (siblings does not include ourself)
440 list_for_each_entry_safe(sibling, next_sibling,
441 &match->rt6i_siblings, rt6i_siblings) {
443 if (route_choosen == 0) {
444 if (rt6_score_route(sibling, oif, strict) < 0)
454 * Route lookup. Any table->tb6_lock is implied.
457 static inline struct rt6_info *rt6_device_match(struct net *net,
459 const struct in6_addr *saddr,
463 struct rt6_info *local = NULL;
464 struct rt6_info *sprt;
466 if (!oif && ipv6_addr_any(saddr))
469 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
470 struct net_device *dev = sprt->dst.dev;
473 if (dev->ifindex == oif)
475 if (dev->flags & IFF_LOOPBACK) {
476 if (!sprt->rt6i_idev ||
477 sprt->rt6i_idev->dev->ifindex != oif) {
478 if (flags & RT6_LOOKUP_F_IFACE)
481 local->rt6i_idev->dev->ifindex == oif)
487 if (ipv6_chk_addr(net, saddr, dev,
488 flags & RT6_LOOKUP_F_IFACE))
497 if (flags & RT6_LOOKUP_F_IFACE)
498 return net->ipv6.ip6_null_entry;
504 #ifdef CONFIG_IPV6_ROUTER_PREF
505 struct __rt6_probe_work {
506 struct work_struct work;
507 struct in6_addr target;
508 struct net_device *dev;
511 static void rt6_probe_deferred(struct work_struct *w)
513 struct in6_addr mcaddr;
514 struct __rt6_probe_work *work =
515 container_of(w, struct __rt6_probe_work, work);
517 addrconf_addr_solict_mult(&work->target, &mcaddr);
518 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, NULL);
523 static void rt6_probe(struct rt6_info *rt)
525 struct __rt6_probe_work *work;
526 struct neighbour *neigh;
528 * Okay, this does not seem to be appropriate
529 * for now, however, we need to check if it
530 * is really so; aka Router Reachability Probing.
532 * Router Reachability Probe MUST be rate-limited
533 * to no more than one per minute.
535 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
538 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
540 if (neigh->nud_state & NUD_VALID)
544 write_lock(&neigh->lock);
545 if (!(neigh->nud_state & NUD_VALID) &&
548 rt->rt6i_idev->cnf.rtr_probe_interval)) {
549 work = kmalloc(sizeof(*work), GFP_ATOMIC);
551 __neigh_set_probe_once(neigh);
553 write_unlock(&neigh->lock);
555 work = kmalloc(sizeof(*work), GFP_ATOMIC);
559 INIT_WORK(&work->work, rt6_probe_deferred);
560 work->target = rt->rt6i_gateway;
561 dev_hold(rt->dst.dev);
562 work->dev = rt->dst.dev;
563 schedule_work(&work->work);
567 rcu_read_unlock_bh();
570 static inline void rt6_probe(struct rt6_info *rt)
576 * Default Router Selection (RFC 2461 6.3.6)
578 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
580 struct net_device *dev = rt->dst.dev;
581 if (!oif || dev->ifindex == oif)
583 if ((dev->flags & IFF_LOOPBACK) &&
584 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
589 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
591 struct neighbour *neigh;
592 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
594 if (rt->rt6i_flags & RTF_NONEXTHOP ||
595 !(rt->rt6i_flags & RTF_GATEWAY))
596 return RT6_NUD_SUCCEED;
599 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
601 read_lock(&neigh->lock);
602 if (neigh->nud_state & NUD_VALID)
603 ret = RT6_NUD_SUCCEED;
604 #ifdef CONFIG_IPV6_ROUTER_PREF
605 else if (!(neigh->nud_state & NUD_FAILED))
606 ret = RT6_NUD_SUCCEED;
608 ret = RT6_NUD_FAIL_PROBE;
610 read_unlock(&neigh->lock);
612 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
613 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
615 rcu_read_unlock_bh();
620 static int rt6_score_route(struct rt6_info *rt, int oif,
625 m = rt6_check_dev(rt, oif);
626 if (!m && (strict & RT6_LOOKUP_F_IFACE))
627 return RT6_NUD_FAIL_HARD;
628 #ifdef CONFIG_IPV6_ROUTER_PREF
629 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
631 if (strict & RT6_LOOKUP_F_REACHABLE) {
632 int n = rt6_check_neigh(rt);
639 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
640 int *mpri, struct rt6_info *match,
644 bool match_do_rr = false;
645 struct inet6_dev *idev = rt->rt6i_idev;
646 struct net_device *dev = rt->dst.dev;
648 if (dev && !netif_carrier_ok(dev) &&
649 idev->cnf.ignore_routes_with_linkdown)
652 if (rt6_check_expired(rt))
655 m = rt6_score_route(rt, oif, strict);
656 if (m == RT6_NUD_FAIL_DO_RR) {
658 m = 0; /* lowest valid score */
659 } else if (m == RT6_NUD_FAIL_HARD) {
663 if (strict & RT6_LOOKUP_F_REACHABLE)
666 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
668 *do_rr = match_do_rr;
676 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
677 struct rt6_info *rr_head,
678 u32 metric, int oif, int strict,
681 struct rt6_info *rt, *match, *cont;
686 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
687 if (rt->rt6i_metric != metric) {
692 match = find_match(rt, oif, strict, &mpri, match, do_rr);
695 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
696 if (rt->rt6i_metric != metric) {
701 match = find_match(rt, oif, strict, &mpri, match, do_rr);
707 for (rt = cont; rt; rt = rt->dst.rt6_next)
708 match = find_match(rt, oif, strict, &mpri, match, do_rr);
713 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
715 struct rt6_info *match, *rt0;
721 fn->rr_ptr = rt0 = fn->leaf;
723 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
727 struct rt6_info *next = rt0->dst.rt6_next;
729 /* no entries matched; do round-robin */
730 if (!next || next->rt6i_metric != rt0->rt6i_metric)
737 net = dev_net(rt0->dst.dev);
738 return match ? match : net->ipv6.ip6_null_entry;
741 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
743 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
746 #ifdef CONFIG_IPV6_ROUTE_INFO
747 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
748 const struct in6_addr *gwaddr)
750 struct net *net = dev_net(dev);
751 struct route_info *rinfo = (struct route_info *) opt;
752 struct in6_addr prefix_buf, *prefix;
754 unsigned long lifetime;
757 if (len < sizeof(struct route_info)) {
761 /* Sanity check for prefix_len and length */
762 if (rinfo->length > 3) {
764 } else if (rinfo->prefix_len > 128) {
766 } else if (rinfo->prefix_len > 64) {
767 if (rinfo->length < 2) {
770 } else if (rinfo->prefix_len > 0) {
771 if (rinfo->length < 1) {
776 pref = rinfo->route_pref;
777 if (pref == ICMPV6_ROUTER_PREF_INVALID)
780 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
782 if (rinfo->length == 3)
783 prefix = (struct in6_addr *)rinfo->prefix;
785 /* this function is safe */
786 ipv6_addr_prefix(&prefix_buf,
787 (struct in6_addr *)rinfo->prefix,
789 prefix = &prefix_buf;
792 if (rinfo->prefix_len == 0)
793 rt = rt6_get_dflt_router(gwaddr, dev);
795 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
796 gwaddr, dev->ifindex);
798 if (rt && !lifetime) {
804 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
807 rt->rt6i_flags = RTF_ROUTEINFO |
808 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
811 if (!addrconf_finite_timeout(lifetime))
812 rt6_clean_expires(rt);
814 rt6_set_expires(rt, jiffies + HZ * lifetime);
822 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
823 struct in6_addr *saddr)
825 struct fib6_node *pn;
827 if (fn->fn_flags & RTN_TL_ROOT)
830 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
831 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
834 if (fn->fn_flags & RTN_RTINFO)
839 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
840 struct fib6_table *table,
841 struct flowi6 *fl6, int flags)
843 struct fib6_node *fn;
846 read_lock_bh(&table->tb6_lock);
847 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
850 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
851 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
852 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
853 if (rt == net->ipv6.ip6_null_entry) {
854 fn = fib6_backtrack(fn, &fl6->saddr);
858 dst_use(&rt->dst, jiffies);
859 read_unlock_bh(&table->tb6_lock);
864 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
867 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
869 EXPORT_SYMBOL_GPL(ip6_route_lookup);
871 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
872 const struct in6_addr *saddr, int oif, int strict)
874 struct flowi6 fl6 = {
878 struct dst_entry *dst;
879 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
882 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
883 flags |= RT6_LOOKUP_F_HAS_SADDR;
886 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
888 return (struct rt6_info *) dst;
894 EXPORT_SYMBOL(rt6_lookup);
896 /* ip6_ins_rt is called with FREE table->tb6_lock.
897 It takes new route entry, the addition fails by any reason the
898 route is freed. In any case, if caller does not hold it, it may
902 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
903 struct mx6_config *mxc)
906 struct fib6_table *table;
908 table = rt->rt6i_table;
909 write_lock_bh(&table->tb6_lock);
910 err = fib6_add(&table->tb6_root, rt, info, mxc);
911 write_unlock_bh(&table->tb6_lock);
916 int ip6_ins_rt(struct rt6_info *rt)
918 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
919 struct mx6_config mxc = { .mx = NULL, };
921 return __ip6_ins_rt(rt, &info, &mxc);
924 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
925 const struct in6_addr *daddr,
926 const struct in6_addr *saddr)
934 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
935 ort = (struct rt6_info *)ort->dst.from;
937 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
942 ip6_rt_copy_init(rt, ort);
943 rt->rt6i_flags |= RTF_CACHE;
945 rt->dst.flags |= DST_HOST;
946 rt->rt6i_dst.addr = *daddr;
947 rt->rt6i_dst.plen = 128;
949 if (!rt6_is_gw_or_nonexthop(ort)) {
950 if (ort->rt6i_dst.plen != 128 &&
951 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
952 rt->rt6i_flags |= RTF_ANYCAST;
953 #ifdef CONFIG_IPV6_SUBTREES
954 if (rt->rt6i_src.plen && saddr) {
955 rt->rt6i_src.addr = *saddr;
956 rt->rt6i_src.plen = 128;
964 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
966 struct rt6_info *pcpu_rt;
968 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
969 rt->dst.dev, rt->dst.flags);
973 ip6_rt_copy_init(pcpu_rt, rt);
974 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
975 pcpu_rt->rt6i_flags |= RTF_PCPU;
979 /* It should be called with read_lock_bh(&tb6_lock) acquired */
980 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
982 struct rt6_info *pcpu_rt, **p;
984 p = this_cpu_ptr(rt->rt6i_pcpu);
988 dst_hold(&pcpu_rt->dst);
989 rt6_dst_from_metrics_check(pcpu_rt);
994 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
996 struct fib6_table *table = rt->rt6i_table;
997 struct rt6_info *pcpu_rt, *prev, **p;
999 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1001 struct net *net = dev_net(rt->dst.dev);
1003 dst_hold(&net->ipv6.ip6_null_entry->dst);
1004 return net->ipv6.ip6_null_entry;
1007 read_lock_bh(&table->tb6_lock);
1008 if (rt->rt6i_pcpu) {
1009 p = this_cpu_ptr(rt->rt6i_pcpu);
1010 prev = cmpxchg(p, NULL, pcpu_rt);
1012 /* If someone did it before us, return prev instead */
1013 dst_destroy(&pcpu_rt->dst);
1017 /* rt has been removed from the fib6 tree
1018 * before we have a chance to acquire the read_lock.
1019 * In this case, don't brother to create a pcpu rt
1020 * since rt is going away anyway. The next
1021 * dst_check() will trigger a re-lookup.
1023 dst_destroy(&pcpu_rt->dst);
1026 dst_hold(&pcpu_rt->dst);
1027 rt6_dst_from_metrics_check(pcpu_rt);
1028 read_unlock_bh(&table->tb6_lock);
1032 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1033 struct flowi6 *fl6, int flags)
1035 struct fib6_node *fn, *saved_fn;
1036 struct rt6_info *rt;
1039 strict |= flags & RT6_LOOKUP_F_IFACE;
1040 if (net->ipv6.devconf_all->forwarding == 0)
1041 strict |= RT6_LOOKUP_F_REACHABLE;
1043 read_lock_bh(&table->tb6_lock);
1045 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1048 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1052 rt = rt6_select(fn, oif, strict);
1053 if (rt->rt6i_nsiblings)
1054 rt = rt6_multipath_select(rt, fl6, oif, strict);
1055 if (rt == net->ipv6.ip6_null_entry) {
1056 fn = fib6_backtrack(fn, &fl6->saddr);
1058 goto redo_rt6_select;
1059 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1060 /* also consider unreachable route */
1061 strict &= ~RT6_LOOKUP_F_REACHABLE;
1063 goto redo_rt6_select;
1068 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1069 dst_use(&rt->dst, jiffies);
1070 read_unlock_bh(&table->tb6_lock);
1072 rt6_dst_from_metrics_check(rt);
1074 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1075 !(rt->rt6i_flags & RTF_GATEWAY))) {
1076 /* Create a RTF_CACHE clone which will not be
1077 * owned by the fib6 tree. It is for the special case where
1078 * the daddr in the skb during the neighbor look-up is different
1079 * from the fl6->daddr used to look-up route here.
1082 struct rt6_info *uncached_rt;
1084 dst_use(&rt->dst, jiffies);
1085 read_unlock_bh(&table->tb6_lock);
1087 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1088 dst_release(&rt->dst);
1091 rt6_uncached_list_add(uncached_rt);
1093 uncached_rt = net->ipv6.ip6_null_entry;
1095 dst_hold(&uncached_rt->dst);
1099 /* Get a percpu copy */
1101 struct rt6_info *pcpu_rt;
1103 rt->dst.lastuse = jiffies;
1105 pcpu_rt = rt6_get_pcpu_route(rt);
1108 read_unlock_bh(&table->tb6_lock);
1110 /* We have to do the read_unlock first
1111 * because rt6_make_pcpu_route() may trigger
1112 * ip6_dst_gc() which will take the write_lock.
1115 read_unlock_bh(&table->tb6_lock);
1116 pcpu_rt = rt6_make_pcpu_route(rt);
1117 dst_release(&rt->dst);
1125 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1126 struct flowi6 *fl6, int flags)
1128 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1131 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1132 struct net_device *dev,
1133 struct flowi6 *fl6, int flags)
1135 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1136 flags |= RT6_LOOKUP_F_IFACE;
1138 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1141 void ip6_route_input(struct sk_buff *skb)
1143 const struct ipv6hdr *iph = ipv6_hdr(skb);
1144 struct net *net = dev_net(skb->dev);
1145 int flags = RT6_LOOKUP_F_HAS_SADDR;
1146 struct ip_tunnel_info *tun_info;
1147 struct flowi6 fl6 = {
1148 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1149 .daddr = iph->daddr,
1150 .saddr = iph->saddr,
1151 .flowlabel = ip6_flowinfo(iph),
1152 .flowi6_mark = skb->mark,
1153 .flowi6_proto = iph->nexthdr,
1156 tun_info = skb_tunnel_info(skb);
1157 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1158 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1160 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1163 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1164 struct flowi6 *fl6, int flags)
1166 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1169 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1172 struct dst_entry *dst;
1175 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1179 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1181 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1183 flags |= RT6_LOOKUP_F_IFACE;
1185 if (!ipv6_addr_any(&fl6->saddr))
1186 flags |= RT6_LOOKUP_F_HAS_SADDR;
1188 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1190 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1192 EXPORT_SYMBOL(ip6_route_output);
1194 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1196 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1197 struct dst_entry *new = NULL;
1199 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1203 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1206 new->input = dst_discard;
1207 new->output = dst_discard_out;
1209 if (dst_metrics_read_only(&ort->dst))
1210 new->_metrics = ort->dst._metrics;
1212 dst_copy_metrics(new, &ort->dst);
1213 rt->rt6i_idev = ort->rt6i_idev;
1215 in6_dev_hold(rt->rt6i_idev);
1217 rt->rt6i_gateway = ort->rt6i_gateway;
1218 rt->rt6i_flags = ort->rt6i_flags;
1219 rt->rt6i_metric = 0;
1221 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1222 #ifdef CONFIG_IPV6_SUBTREES
1223 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1229 dst_release(dst_orig);
1230 return new ? new : ERR_PTR(-ENOMEM);
1234 * Destination cache support functions
1237 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1240 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1241 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1244 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1246 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1249 if (rt6_check_expired(rt))
1255 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1257 if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1258 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1264 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1266 struct rt6_info *rt;
1268 rt = (struct rt6_info *) dst;
1270 /* All IPV6 dsts are created with ->obsolete set to the value
1271 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1272 * into this function always.
1275 rt6_dst_from_metrics_check(rt);
1277 if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1278 return rt6_dst_from_check(rt, cookie);
1280 return rt6_check(rt, cookie);
1283 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1285 struct rt6_info *rt = (struct rt6_info *) dst;
1288 if (rt->rt6i_flags & RTF_CACHE) {
1289 if (rt6_check_expired(rt)) {
1301 static void ip6_link_failure(struct sk_buff *skb)
1303 struct rt6_info *rt;
1305 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1307 rt = (struct rt6_info *) skb_dst(skb);
1309 if (rt->rt6i_flags & RTF_CACHE) {
1312 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1313 rt->rt6i_node->fn_sernum = -1;
1318 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1320 struct net *net = dev_net(rt->dst.dev);
1322 rt->rt6i_flags |= RTF_MODIFIED;
1323 rt->rt6i_pmtu = mtu;
1324 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1327 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1328 const struct ipv6hdr *iph, u32 mtu)
1330 struct rt6_info *rt6 = (struct rt6_info *)dst;
1332 if (rt6->rt6i_flags & RTF_LOCAL)
1336 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1337 if (mtu >= dst_mtu(dst))
1340 if (rt6->rt6i_flags & RTF_CACHE) {
1341 rt6_do_update_pmtu(rt6, mtu);
1343 const struct in6_addr *daddr, *saddr;
1344 struct rt6_info *nrt6;
1347 daddr = &iph->daddr;
1348 saddr = &iph->saddr;
1350 daddr = &sk->sk_v6_daddr;
1351 saddr = &inet6_sk(sk)->saddr;
1355 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1357 rt6_do_update_pmtu(nrt6, mtu);
1359 /* ip6_ins_rt(nrt6) will bump the
1360 * rt6->rt6i_node->fn_sernum
1361 * which will fail the next rt6_check() and
1362 * invalidate the sk->sk_dst_cache.
1369 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1370 struct sk_buff *skb, u32 mtu)
1372 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1375 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1378 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1379 struct dst_entry *dst;
1382 memset(&fl6, 0, sizeof(fl6));
1383 fl6.flowi6_oif = oif;
1384 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1385 fl6.daddr = iph->daddr;
1386 fl6.saddr = iph->saddr;
1387 fl6.flowlabel = ip6_flowinfo(iph);
1389 dst = ip6_route_output(net, NULL, &fl6);
1391 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1394 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1396 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1398 ip6_update_pmtu(skb, sock_net(sk), mtu,
1399 sk->sk_bound_dev_if, sk->sk_mark);
1401 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1403 /* Handle redirects */
1404 struct ip6rd_flowi {
1406 struct in6_addr gateway;
1409 static struct rt6_info *__ip6_route_redirect(struct net *net,
1410 struct fib6_table *table,
1414 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1415 struct rt6_info *rt;
1416 struct fib6_node *fn;
1418 /* Get the "current" route for this destination and
1419 * check if the redirect has come from approriate router.
1421 * RFC 4861 specifies that redirects should only be
1422 * accepted if they come from the nexthop to the target.
1423 * Due to the way the routes are chosen, this notion
1424 * is a bit fuzzy and one might need to check all possible
1428 read_lock_bh(&table->tb6_lock);
1429 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1431 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1432 if (rt6_check_expired(rt))
1436 if (!(rt->rt6i_flags & RTF_GATEWAY))
1438 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1440 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1446 rt = net->ipv6.ip6_null_entry;
1447 else if (rt->dst.error) {
1448 rt = net->ipv6.ip6_null_entry;
1452 if (rt == net->ipv6.ip6_null_entry) {
1453 fn = fib6_backtrack(fn, &fl6->saddr);
1461 read_unlock_bh(&table->tb6_lock);
1466 static struct dst_entry *ip6_route_redirect(struct net *net,
1467 const struct flowi6 *fl6,
1468 const struct in6_addr *gateway)
1470 int flags = RT6_LOOKUP_F_HAS_SADDR;
1471 struct ip6rd_flowi rdfl;
1474 rdfl.gateway = *gateway;
1476 return fib6_rule_lookup(net, &rdfl.fl6,
1477 flags, __ip6_route_redirect);
1480 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1482 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1483 struct dst_entry *dst;
1486 memset(&fl6, 0, sizeof(fl6));
1487 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1488 fl6.flowi6_oif = oif;
1489 fl6.flowi6_mark = mark;
1490 fl6.daddr = iph->daddr;
1491 fl6.saddr = iph->saddr;
1492 fl6.flowlabel = ip6_flowinfo(iph);
1494 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1495 rt6_do_redirect(dst, NULL, skb);
1498 EXPORT_SYMBOL_GPL(ip6_redirect);
1500 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1503 const struct ipv6hdr *iph = ipv6_hdr(skb);
1504 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1505 struct dst_entry *dst;
1508 memset(&fl6, 0, sizeof(fl6));
1509 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1510 fl6.flowi6_oif = oif;
1511 fl6.flowi6_mark = mark;
1512 fl6.daddr = msg->dest;
1513 fl6.saddr = iph->daddr;
1515 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1516 rt6_do_redirect(dst, NULL, skb);
1520 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1522 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1524 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1526 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1528 struct net_device *dev = dst->dev;
1529 unsigned int mtu = dst_mtu(dst);
1530 struct net *net = dev_net(dev);
1532 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1534 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1535 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1538 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1539 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1540 * IPV6_MAXPLEN is also valid and means: "any MSS,
1541 * rely only on pmtu discovery"
1543 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1548 static unsigned int ip6_mtu(const struct dst_entry *dst)
1550 const struct rt6_info *rt = (const struct rt6_info *)dst;
1551 unsigned int mtu = rt->rt6i_pmtu;
1552 struct inet6_dev *idev;
1557 mtu = dst_metric_raw(dst, RTAX_MTU);
1564 idev = __in6_dev_get(dst->dev);
1566 mtu = idev->cnf.mtu6;
1570 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1573 static struct dst_entry *icmp6_dst_gc_list;
1574 static DEFINE_SPINLOCK(icmp6_dst_lock);
1576 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1579 struct dst_entry *dst;
1580 struct rt6_info *rt;
1581 struct inet6_dev *idev = in6_dev_get(dev);
1582 struct net *net = dev_net(dev);
1584 if (unlikely(!idev))
1585 return ERR_PTR(-ENODEV);
1587 rt = ip6_dst_alloc(net, dev, 0);
1588 if (unlikely(!rt)) {
1590 dst = ERR_PTR(-ENOMEM);
1594 rt->dst.flags |= DST_HOST;
1595 rt->dst.output = ip6_output;
1596 atomic_set(&rt->dst.__refcnt, 1);
1597 rt->rt6i_gateway = fl6->daddr;
1598 rt->rt6i_dst.addr = fl6->daddr;
1599 rt->rt6i_dst.plen = 128;
1600 rt->rt6i_idev = idev;
1601 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1603 spin_lock_bh(&icmp6_dst_lock);
1604 rt->dst.next = icmp6_dst_gc_list;
1605 icmp6_dst_gc_list = &rt->dst;
1606 spin_unlock_bh(&icmp6_dst_lock);
1608 fib6_force_start_gc(net);
1610 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1616 int icmp6_dst_gc(void)
1618 struct dst_entry *dst, **pprev;
1621 spin_lock_bh(&icmp6_dst_lock);
1622 pprev = &icmp6_dst_gc_list;
1624 while ((dst = *pprev) != NULL) {
1625 if (!atomic_read(&dst->__refcnt)) {
1634 spin_unlock_bh(&icmp6_dst_lock);
1639 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1642 struct dst_entry *dst, **pprev;
1644 spin_lock_bh(&icmp6_dst_lock);
1645 pprev = &icmp6_dst_gc_list;
1646 while ((dst = *pprev) != NULL) {
1647 struct rt6_info *rt = (struct rt6_info *) dst;
1648 if (func(rt, arg)) {
1655 spin_unlock_bh(&icmp6_dst_lock);
1658 static int ip6_dst_gc(struct dst_ops *ops)
1660 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1661 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1662 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1663 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1664 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1665 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1668 entries = dst_entries_get_fast(ops);
1669 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1670 entries <= rt_max_size)
1673 net->ipv6.ip6_rt_gc_expire++;
1674 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1675 entries = dst_entries_get_slow(ops);
1676 if (entries < ops->gc_thresh)
1677 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1679 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1680 return entries > rt_max_size;
1683 static int ip6_convert_metrics(struct mx6_config *mxc,
1684 const struct fib6_config *cfg)
1686 bool ecn_ca = false;
1694 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1698 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1699 int type = nla_type(nla);
1704 if (unlikely(type > RTAX_MAX))
1707 if (type == RTAX_CC_ALGO) {
1708 char tmp[TCP_CA_NAME_MAX];
1710 nla_strlcpy(tmp, nla, sizeof(tmp));
1711 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1712 if (val == TCP_CA_UNSPEC)
1715 val = nla_get_u32(nla);
1717 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1721 __set_bit(type - 1, mxc->mx_valid);
1725 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1726 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1736 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1738 struct net *net = cfg->fc_nlinfo.nl_net;
1739 struct rt6_info *rt = NULL;
1740 struct net_device *dev = NULL;
1741 struct inet6_dev *idev = NULL;
1742 struct fib6_table *table;
1746 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1748 #ifndef CONFIG_IPV6_SUBTREES
1749 if (cfg->fc_src_len)
1752 if (cfg->fc_ifindex) {
1754 dev = dev_get_by_index(net, cfg->fc_ifindex);
1757 idev = in6_dev_get(dev);
1762 if (cfg->fc_metric == 0)
1763 cfg->fc_metric = IP6_RT_PRIO_USER;
1766 if (cfg->fc_nlinfo.nlh &&
1767 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1768 table = fib6_get_table(net, cfg->fc_table);
1770 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1771 table = fib6_new_table(net, cfg->fc_table);
1774 table = fib6_new_table(net, cfg->fc_table);
1780 rt = ip6_dst_alloc(net, NULL,
1781 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1788 if (cfg->fc_flags & RTF_EXPIRES)
1789 rt6_set_expires(rt, jiffies +
1790 clock_t_to_jiffies(cfg->fc_expires));
1792 rt6_clean_expires(rt);
1794 if (cfg->fc_protocol == RTPROT_UNSPEC)
1795 cfg->fc_protocol = RTPROT_BOOT;
1796 rt->rt6i_protocol = cfg->fc_protocol;
1798 addr_type = ipv6_addr_type(&cfg->fc_dst);
1800 if (addr_type & IPV6_ADDR_MULTICAST)
1801 rt->dst.input = ip6_mc_input;
1802 else if (cfg->fc_flags & RTF_LOCAL)
1803 rt->dst.input = ip6_input;
1805 rt->dst.input = ip6_forward;
1807 rt->dst.output = ip6_output;
1809 if (cfg->fc_encap) {
1810 struct lwtunnel_state *lwtstate;
1812 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1813 cfg->fc_encap, AF_INET6, cfg,
1817 rt->dst.lwtstate = lwtstate_get(lwtstate);
1818 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1819 rt->dst.lwtstate->orig_output = rt->dst.output;
1820 rt->dst.output = lwtunnel_output;
1822 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1823 rt->dst.lwtstate->orig_input = rt->dst.input;
1824 rt->dst.input = lwtunnel_input;
1828 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1829 rt->rt6i_dst.plen = cfg->fc_dst_len;
1830 if (rt->rt6i_dst.plen == 128)
1831 rt->dst.flags |= DST_HOST;
1833 #ifdef CONFIG_IPV6_SUBTREES
1834 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1835 rt->rt6i_src.plen = cfg->fc_src_len;
1838 rt->rt6i_metric = cfg->fc_metric;
1840 /* We cannot add true routes via loopback here,
1841 they would result in kernel looping; promote them to reject routes
1843 if ((cfg->fc_flags & RTF_REJECT) ||
1844 (dev && (dev->flags & IFF_LOOPBACK) &&
1845 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1846 !(cfg->fc_flags & RTF_LOCAL))) {
1847 /* hold loopback dev/idev if we haven't done so. */
1848 if (dev != net->loopback_dev) {
1853 dev = net->loopback_dev;
1855 idev = in6_dev_get(dev);
1861 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1862 switch (cfg->fc_type) {
1864 rt->dst.error = -EINVAL;
1865 rt->dst.output = dst_discard_out;
1866 rt->dst.input = dst_discard;
1869 rt->dst.error = -EACCES;
1870 rt->dst.output = ip6_pkt_prohibit_out;
1871 rt->dst.input = ip6_pkt_prohibit;
1874 case RTN_UNREACHABLE:
1876 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1877 : (cfg->fc_type == RTN_UNREACHABLE)
1878 ? -EHOSTUNREACH : -ENETUNREACH;
1879 rt->dst.output = ip6_pkt_discard_out;
1880 rt->dst.input = ip6_pkt_discard;
1886 if (cfg->fc_flags & RTF_GATEWAY) {
1887 const struct in6_addr *gw_addr;
1890 gw_addr = &cfg->fc_gateway;
1891 gwa_type = ipv6_addr_type(gw_addr);
1893 /* if gw_addr is local we will fail to detect this in case
1894 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1895 * will return already-added prefix route via interface that
1896 * prefix route was assigned to, which might be non-loopback.
1899 if (ipv6_chk_addr_and_flags(net, gw_addr,
1900 gwa_type & IPV6_ADDR_LINKLOCAL ?
1904 rt->rt6i_gateway = *gw_addr;
1906 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1907 struct rt6_info *grt;
1909 /* IPv6 strictly inhibits using not link-local
1910 addresses as nexthop address.
1911 Otherwise, router will not able to send redirects.
1912 It is very good, but in some (rare!) circumstances
1913 (SIT, PtP, NBMA NOARP links) it is handy to allow
1914 some exceptions. --ANK
1916 if (!(gwa_type & IPV6_ADDR_UNICAST))
1919 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1921 err = -EHOSTUNREACH;
1925 if (dev != grt->dst.dev) {
1931 idev = grt->rt6i_idev;
1933 in6_dev_hold(grt->rt6i_idev);
1935 if (!(grt->rt6i_flags & RTF_GATEWAY))
1943 if (!dev || (dev->flags & IFF_LOOPBACK))
1951 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1952 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1956 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1957 rt->rt6i_prefsrc.plen = 128;
1959 rt->rt6i_prefsrc.plen = 0;
1961 rt->rt6i_flags = cfg->fc_flags;
1965 rt->rt6i_idev = idev;
1966 rt->rt6i_table = table;
1968 cfg->fc_nlinfo.nl_net = dev_net(dev);
1979 return ERR_PTR(err);
1982 int ip6_route_add(struct fib6_config *cfg)
1984 struct mx6_config mxc = { .mx = NULL, };
1985 struct rt6_info *rt;
1988 rt = ip6_route_info_create(cfg);
1995 err = ip6_convert_metrics(&mxc, cfg);
1999 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2011 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2014 struct fib6_table *table;
2015 struct net *net = dev_net(rt->dst.dev);
2017 if (rt == net->ipv6.ip6_null_entry ||
2018 rt->dst.flags & DST_NOCACHE) {
2023 table = rt->rt6i_table;
2024 write_lock_bh(&table->tb6_lock);
2025 err = fib6_del(rt, info);
2026 write_unlock_bh(&table->tb6_lock);
2033 int ip6_del_rt(struct rt6_info *rt)
2035 struct nl_info info = {
2036 .nl_net = dev_net(rt->dst.dev),
2038 return __ip6_del_rt(rt, &info);
2041 static int ip6_route_del(struct fib6_config *cfg)
2043 struct fib6_table *table;
2044 struct fib6_node *fn;
2045 struct rt6_info *rt;
2048 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2052 read_lock_bh(&table->tb6_lock);
2054 fn = fib6_locate(&table->tb6_root,
2055 &cfg->fc_dst, cfg->fc_dst_len,
2056 &cfg->fc_src, cfg->fc_src_len);
2059 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2060 if ((rt->rt6i_flags & RTF_CACHE) &&
2061 !(cfg->fc_flags & RTF_CACHE))
2063 if (cfg->fc_ifindex &&
2065 rt->dst.dev->ifindex != cfg->fc_ifindex))
2067 if (cfg->fc_flags & RTF_GATEWAY &&
2068 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2070 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2073 read_unlock_bh(&table->tb6_lock);
2075 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2078 read_unlock_bh(&table->tb6_lock);
2083 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2085 struct net *net = dev_net(skb->dev);
2086 struct netevent_redirect netevent;
2087 struct rt6_info *rt, *nrt = NULL;
2088 struct ndisc_options ndopts;
2089 struct inet6_dev *in6_dev;
2090 struct neighbour *neigh;
2092 int optlen, on_link;
2095 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2096 optlen -= sizeof(*msg);
2099 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2103 msg = (struct rd_msg *)icmp6_hdr(skb);
2105 if (ipv6_addr_is_multicast(&msg->dest)) {
2106 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2111 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2113 } else if (ipv6_addr_type(&msg->target) !=
2114 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2115 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2119 in6_dev = __in6_dev_get(skb->dev);
2122 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2126 * The IP source address of the Redirect MUST be the same as the current
2127 * first-hop router for the specified ICMP Destination Address.
2130 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2131 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2136 if (ndopts.nd_opts_tgt_lladdr) {
2137 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2140 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2145 rt = (struct rt6_info *) dst;
2146 if (rt == net->ipv6.ip6_null_entry) {
2147 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2151 /* Redirect received -> path was valid.
2152 * Look, redirects are sent only in response to data packets,
2153 * so that this nexthop apparently is reachable. --ANK
2155 dst_confirm(&rt->dst);
2157 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2162 * We have finally decided to accept it.
2165 neigh_update(neigh, lladdr, NUD_STALE,
2166 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2167 NEIGH_UPDATE_F_OVERRIDE|
2168 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2169 NEIGH_UPDATE_F_ISROUTER))
2172 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2176 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2178 nrt->rt6i_flags &= ~RTF_GATEWAY;
2180 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2182 if (ip6_ins_rt(nrt))
2185 netevent.old = &rt->dst;
2186 netevent.new = &nrt->dst;
2187 netevent.daddr = &msg->dest;
2188 netevent.neigh = neigh;
2189 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2191 if (rt->rt6i_flags & RTF_CACHE) {
2192 rt = (struct rt6_info *) dst_clone(&rt->dst);
2197 neigh_release(neigh);
2201 * Misc support functions
2204 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2206 BUG_ON(from->dst.from);
2208 rt->rt6i_flags &= ~RTF_EXPIRES;
2209 dst_hold(&from->dst);
2210 rt->dst.from = &from->dst;
2211 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2214 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2216 rt->dst.input = ort->dst.input;
2217 rt->dst.output = ort->dst.output;
2218 rt->rt6i_dst = ort->rt6i_dst;
2219 rt->dst.error = ort->dst.error;
2220 rt->rt6i_idev = ort->rt6i_idev;
2222 in6_dev_hold(rt->rt6i_idev);
2223 rt->dst.lastuse = jiffies;
2224 rt->rt6i_gateway = ort->rt6i_gateway;
2225 rt->rt6i_flags = ort->rt6i_flags;
2226 rt6_set_from(rt, ort);
2227 rt->rt6i_metric = ort->rt6i_metric;
2228 #ifdef CONFIG_IPV6_SUBTREES
2229 rt->rt6i_src = ort->rt6i_src;
2231 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2232 rt->rt6i_table = ort->rt6i_table;
2233 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2236 #ifdef CONFIG_IPV6_ROUTE_INFO
2237 static struct rt6_info *rt6_get_route_info(struct net *net,
2238 const struct in6_addr *prefix, int prefixlen,
2239 const struct in6_addr *gwaddr, int ifindex)
2241 struct fib6_node *fn;
2242 struct rt6_info *rt = NULL;
2243 struct fib6_table *table;
2245 table = fib6_get_table(net, RT6_TABLE_INFO);
2249 read_lock_bh(&table->tb6_lock);
2250 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2254 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2255 if (rt->dst.dev->ifindex != ifindex)
2257 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2259 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2265 read_unlock_bh(&table->tb6_lock);
2269 static struct rt6_info *rt6_add_route_info(struct net *net,
2270 const struct in6_addr *prefix, int prefixlen,
2271 const struct in6_addr *gwaddr, int ifindex,
2274 struct fib6_config cfg = {
2275 .fc_metric = IP6_RT_PRIO_USER,
2276 .fc_ifindex = ifindex,
2277 .fc_dst_len = prefixlen,
2278 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2279 RTF_UP | RTF_PREF(pref),
2280 .fc_nlinfo.portid = 0,
2281 .fc_nlinfo.nlh = NULL,
2282 .fc_nlinfo.nl_net = net,
2285 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2286 cfg.fc_dst = *prefix;
2287 cfg.fc_gateway = *gwaddr;
2289 /* We should treat it as a default route if prefix length is 0. */
2291 cfg.fc_flags |= RTF_DEFAULT;
2293 ip6_route_add(&cfg);
2295 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2299 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2301 struct rt6_info *rt;
2302 struct fib6_table *table;
2304 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2308 read_lock_bh(&table->tb6_lock);
2309 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2310 if (dev == rt->dst.dev &&
2311 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2312 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2317 read_unlock_bh(&table->tb6_lock);
2321 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2322 struct net_device *dev,
2325 struct fib6_config cfg = {
2326 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2327 .fc_metric = IP6_RT_PRIO_USER,
2328 .fc_ifindex = dev->ifindex,
2329 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2330 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2331 .fc_nlinfo.portid = 0,
2332 .fc_nlinfo.nlh = NULL,
2333 .fc_nlinfo.nl_net = dev_net(dev),
2336 cfg.fc_gateway = *gwaddr;
2338 ip6_route_add(&cfg);
2340 return rt6_get_dflt_router(gwaddr, dev);
2343 void rt6_purge_dflt_routers(struct net *net)
2345 struct rt6_info *rt;
2346 struct fib6_table *table;
2348 /* NOTE: Keep consistent with rt6_get_dflt_router */
2349 table = fib6_get_table(net, RT6_TABLE_DFLT);
2354 read_lock_bh(&table->tb6_lock);
2355 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2356 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2357 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2359 read_unlock_bh(&table->tb6_lock);
2364 read_unlock_bh(&table->tb6_lock);
2367 static void rtmsg_to_fib6_config(struct net *net,
2368 struct in6_rtmsg *rtmsg,
2369 struct fib6_config *cfg)
2371 memset(cfg, 0, sizeof(*cfg));
2373 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2375 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2376 cfg->fc_metric = rtmsg->rtmsg_metric;
2377 cfg->fc_expires = rtmsg->rtmsg_info;
2378 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2379 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2380 cfg->fc_flags = rtmsg->rtmsg_flags;
2382 cfg->fc_nlinfo.nl_net = net;
2384 cfg->fc_dst = rtmsg->rtmsg_dst;
2385 cfg->fc_src = rtmsg->rtmsg_src;
2386 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2389 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2391 struct fib6_config cfg;
2392 struct in6_rtmsg rtmsg;
2396 case SIOCADDRT: /* Add a route */
2397 case SIOCDELRT: /* Delete a route */
2398 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2400 err = copy_from_user(&rtmsg, arg,
2401 sizeof(struct in6_rtmsg));
2405 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2410 err = ip6_route_add(&cfg);
2413 err = ip6_route_del(&cfg);
2427 * Drop the packet on the floor
2430 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2433 struct dst_entry *dst = skb_dst(skb);
2434 switch (ipstats_mib_noroutes) {
2435 case IPSTATS_MIB_INNOROUTES:
2436 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2437 if (type == IPV6_ADDR_ANY) {
2438 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2439 IPSTATS_MIB_INADDRERRORS);
2443 case IPSTATS_MIB_OUTNOROUTES:
2444 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2445 ipstats_mib_noroutes);
2448 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2453 static int ip6_pkt_discard(struct sk_buff *skb)
2455 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2458 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2460 skb->dev = skb_dst(skb)->dev;
2461 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2464 static int ip6_pkt_prohibit(struct sk_buff *skb)
2466 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2469 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2471 skb->dev = skb_dst(skb)->dev;
2472 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2476 * Allocate a dst for local (unicast / anycast) address.
2479 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2480 const struct in6_addr *addr,
2484 struct net *net = dev_net(idev->dev);
2485 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2488 return ERR_PTR(-ENOMEM);
2492 rt->dst.flags |= DST_HOST;
2493 rt->dst.input = ip6_input;
2494 rt->dst.output = ip6_output;
2495 rt->rt6i_idev = idev;
2497 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2499 rt->rt6i_flags |= RTF_ANYCAST;
2501 rt->rt6i_flags |= RTF_LOCAL;
2503 rt->rt6i_gateway = *addr;
2504 rt->rt6i_dst.addr = *addr;
2505 rt->rt6i_dst.plen = 128;
2506 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2507 rt->rt6i_table = fib6_get_table(net, tb_id);
2508 rt->dst.flags |= DST_NOCACHE;
2510 atomic_set(&rt->dst.__refcnt, 1);
2515 int ip6_route_get_saddr(struct net *net,
2516 struct rt6_info *rt,
2517 const struct in6_addr *daddr,
2519 struct in6_addr *saddr)
2521 struct inet6_dev *idev =
2522 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2524 if (rt && rt->rt6i_prefsrc.plen)
2525 *saddr = rt->rt6i_prefsrc.addr;
2527 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2528 daddr, prefs, saddr);
2532 /* remove deleted ip from prefsrc entries */
2533 struct arg_dev_net_ip {
2534 struct net_device *dev;
2536 struct in6_addr *addr;
2539 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2541 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2542 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2543 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2545 if (((void *)rt->dst.dev == dev || !dev) &&
2546 rt != net->ipv6.ip6_null_entry &&
2547 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2548 /* remove prefsrc entry */
2549 rt->rt6i_prefsrc.plen = 0;
2554 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2556 struct net *net = dev_net(ifp->idev->dev);
2557 struct arg_dev_net_ip adni = {
2558 .dev = ifp->idev->dev,
2562 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2565 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2566 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2568 /* Remove routers and update dst entries when gateway turn into host. */
2569 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2571 struct in6_addr *gateway = (struct in6_addr *)arg;
2573 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2574 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2575 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2581 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2583 fib6_clean_all(net, fib6_clean_tohost, gateway);
2586 struct arg_dev_net {
2587 struct net_device *dev;
2591 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2593 const struct arg_dev_net *adn = arg;
2594 const struct net_device *dev = adn->dev;
2596 if ((rt->dst.dev == dev || !dev) &&
2597 rt != adn->net->ipv6.ip6_null_entry)
2603 void rt6_ifdown(struct net *net, struct net_device *dev)
2605 struct arg_dev_net adn = {
2610 fib6_clean_all(net, fib6_ifdown, &adn);
2611 icmp6_clean_all(fib6_ifdown, &adn);
2612 rt6_uncached_list_flush_dev(net, dev);
2615 struct rt6_mtu_change_arg {
2616 struct net_device *dev;
2620 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2622 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2623 struct inet6_dev *idev;
2625 /* In IPv6 pmtu discovery is not optional,
2626 so that RTAX_MTU lock cannot disable it.
2627 We still use this lock to block changes
2628 caused by addrconf/ndisc.
2631 idev = __in6_dev_get(arg->dev);
2635 /* For administrative MTU increase, there is no way to discover
2636 IPv6 PMTU increase, so PMTU increase should be updated here.
2637 Since RFC 1981 doesn't include administrative MTU increase
2638 update PMTU increase is a MUST. (i.e. jumbo frame)
2641 If new MTU is less than route PMTU, this new MTU will be the
2642 lowest MTU in the path, update the route PMTU to reflect PMTU
2643 decreases; if new MTU is greater than route PMTU, and the
2644 old MTU is the lowest MTU in the path, update the route PMTU
2645 to reflect the increase. In this case if the other nodes' MTU
2646 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2649 if (rt->dst.dev == arg->dev &&
2650 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2651 if (rt->rt6i_flags & RTF_CACHE) {
2652 /* For RTF_CACHE with rt6i_pmtu == 0
2653 * (i.e. a redirected route),
2654 * the metrics of its rt->dst.from has already
2657 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2658 rt->rt6i_pmtu = arg->mtu;
2659 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2660 (dst_mtu(&rt->dst) < arg->mtu &&
2661 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2662 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2668 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2670 struct rt6_mtu_change_arg arg = {
2675 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2678 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2679 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2680 [RTA_OIF] = { .type = NLA_U32 },
2681 [RTA_IIF] = { .type = NLA_U32 },
2682 [RTA_PRIORITY] = { .type = NLA_U32 },
2683 [RTA_METRICS] = { .type = NLA_NESTED },
2684 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2685 [RTA_PREF] = { .type = NLA_U8 },
2686 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2687 [RTA_ENCAP] = { .type = NLA_NESTED },
2690 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2691 struct fib6_config *cfg)
2694 struct nlattr *tb[RTA_MAX+1];
2698 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2703 rtm = nlmsg_data(nlh);
2704 memset(cfg, 0, sizeof(*cfg));
2706 cfg->fc_table = rtm->rtm_table;
2707 cfg->fc_dst_len = rtm->rtm_dst_len;
2708 cfg->fc_src_len = rtm->rtm_src_len;
2709 cfg->fc_flags = RTF_UP;
2710 cfg->fc_protocol = rtm->rtm_protocol;
2711 cfg->fc_type = rtm->rtm_type;
2713 if (rtm->rtm_type == RTN_UNREACHABLE ||
2714 rtm->rtm_type == RTN_BLACKHOLE ||
2715 rtm->rtm_type == RTN_PROHIBIT ||
2716 rtm->rtm_type == RTN_THROW)
2717 cfg->fc_flags |= RTF_REJECT;
2719 if (rtm->rtm_type == RTN_LOCAL)
2720 cfg->fc_flags |= RTF_LOCAL;
2722 if (rtm->rtm_flags & RTM_F_CLONED)
2723 cfg->fc_flags |= RTF_CACHE;
2725 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2726 cfg->fc_nlinfo.nlh = nlh;
2727 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2729 if (tb[RTA_GATEWAY]) {
2730 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2731 cfg->fc_flags |= RTF_GATEWAY;
2735 int plen = (rtm->rtm_dst_len + 7) >> 3;
2737 if (nla_len(tb[RTA_DST]) < plen)
2740 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2744 int plen = (rtm->rtm_src_len + 7) >> 3;
2746 if (nla_len(tb[RTA_SRC]) < plen)
2749 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2752 if (tb[RTA_PREFSRC])
2753 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2756 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2758 if (tb[RTA_PRIORITY])
2759 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2761 if (tb[RTA_METRICS]) {
2762 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2763 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2767 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2769 if (tb[RTA_MULTIPATH]) {
2770 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2771 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2775 pref = nla_get_u8(tb[RTA_PREF]);
2776 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2777 pref != ICMPV6_ROUTER_PREF_HIGH)
2778 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2779 cfg->fc_flags |= RTF_PREF(pref);
2783 cfg->fc_encap = tb[RTA_ENCAP];
2785 if (tb[RTA_ENCAP_TYPE])
2786 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2794 struct rt6_info *rt6_info;
2795 struct fib6_config r_cfg;
2796 struct mx6_config mxc;
2797 struct list_head next;
2800 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2804 list_for_each_entry(nh, rt6_nh_list, next) {
2805 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2806 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2807 nh->r_cfg.fc_ifindex);
2811 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2812 struct rt6_info *rt, struct fib6_config *r_cfg)
2815 struct rt6_info *rtnh;
2818 list_for_each_entry(nh, rt6_nh_list, next) {
2819 /* check if rt6_info already exists */
2820 rtnh = nh->rt6_info;
2822 if (rtnh->dst.dev == rt->dst.dev &&
2823 rtnh->rt6i_idev == rt->rt6i_idev &&
2824 ipv6_addr_equal(&rtnh->rt6i_gateway,
2829 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2833 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2838 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2839 list_add_tail(&nh->next, rt6_nh_list);
2844 static int ip6_route_multipath_add(struct fib6_config *cfg)
2846 struct fib6_config r_cfg;
2847 struct rtnexthop *rtnh;
2848 struct rt6_info *rt;
2849 struct rt6_nh *err_nh;
2850 struct rt6_nh *nh, *nh_safe;
2855 int replace = (cfg->fc_nlinfo.nlh &&
2856 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2857 LIST_HEAD(rt6_nh_list);
2859 remaining = cfg->fc_mp_len;
2860 rtnh = (struct rtnexthop *)cfg->fc_mp;
2862 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2863 * rt6_info structs per nexthop
2865 while (rtnh_ok(rtnh, remaining)) {
2866 memcpy(&r_cfg, cfg, sizeof(*cfg));
2867 if (rtnh->rtnh_ifindex)
2868 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2870 attrlen = rtnh_attrlen(rtnh);
2872 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2874 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2876 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2877 r_cfg.fc_flags |= RTF_GATEWAY;
2879 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2880 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2882 r_cfg.fc_encap_type = nla_get_u16(nla);
2885 rt = ip6_route_info_create(&r_cfg);
2892 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2898 rtnh = rtnh_next(rtnh, &remaining);
2902 list_for_each_entry(nh, &rt6_nh_list, next) {
2903 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2904 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2905 nh->rt6_info = NULL;
2908 ip6_print_replace_route_err(&rt6_nh_list);
2913 /* Because each route is added like a single route we remove
2914 * these flags after the first nexthop: if there is a collision,
2915 * we have already failed to add the first nexthop:
2916 * fib6_add_rt2node() has rejected it; when replacing, old
2917 * nexthops have been replaced by first new, the rest should
2920 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2928 /* Delete routes that were already added */
2929 list_for_each_entry(nh, &rt6_nh_list, next) {
2932 ip6_route_del(&nh->r_cfg);
2936 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2938 dst_free(&nh->rt6_info->dst);
2940 list_del(&nh->next);
2947 static int ip6_route_multipath_del(struct fib6_config *cfg)
2949 struct fib6_config r_cfg;
2950 struct rtnexthop *rtnh;
2953 int err = 1, last_err = 0;
2955 remaining = cfg->fc_mp_len;
2956 rtnh = (struct rtnexthop *)cfg->fc_mp;
2958 /* Parse a Multipath Entry */
2959 while (rtnh_ok(rtnh, remaining)) {
2960 memcpy(&r_cfg, cfg, sizeof(*cfg));
2961 if (rtnh->rtnh_ifindex)
2962 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2964 attrlen = rtnh_attrlen(rtnh);
2966 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2968 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2970 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2971 r_cfg.fc_flags |= RTF_GATEWAY;
2974 err = ip6_route_del(&r_cfg);
2978 rtnh = rtnh_next(rtnh, &remaining);
2984 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2986 struct fib6_config cfg;
2989 err = rtm_to_fib6_config(skb, nlh, &cfg);
2994 return ip6_route_multipath_del(&cfg);
2996 return ip6_route_del(&cfg);
2999 static int inet6_rtm_newroute(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_add(&cfg);
3011 return ip6_route_add(&cfg);
3014 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3016 return NLMSG_ALIGN(sizeof(struct rtmsg))
3017 + nla_total_size(16) /* RTA_SRC */
3018 + nla_total_size(16) /* RTA_DST */
3019 + nla_total_size(16) /* RTA_GATEWAY */
3020 + nla_total_size(16) /* RTA_PREFSRC */
3021 + nla_total_size(4) /* RTA_TABLE */
3022 + nla_total_size(4) /* RTA_IIF */
3023 + nla_total_size(4) /* RTA_OIF */
3024 + nla_total_size(4) /* RTA_PRIORITY */
3025 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3026 + nla_total_size(sizeof(struct rta_cacheinfo))
3027 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3028 + nla_total_size(1) /* RTA_PREF */
3029 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3032 static int rt6_fill_node(struct net *net,
3033 struct sk_buff *skb, struct rt6_info *rt,
3034 struct in6_addr *dst, struct in6_addr *src,
3035 int iif, int type, u32 portid, u32 seq,
3036 int prefix, int nowait, unsigned int flags)
3038 u32 metrics[RTAX_MAX];
3040 struct nlmsghdr *nlh;
3044 if (prefix) { /* user wants prefix routes only */
3045 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3046 /* success since this is not a prefix route */
3051 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3055 rtm = nlmsg_data(nlh);
3056 rtm->rtm_family = AF_INET6;
3057 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3058 rtm->rtm_src_len = rt->rt6i_src.plen;
3061 table = rt->rt6i_table->tb6_id;
3063 table = RT6_TABLE_UNSPEC;
3064 rtm->rtm_table = table;
3065 if (nla_put_u32(skb, RTA_TABLE, table))
3066 goto nla_put_failure;
3067 if (rt->rt6i_flags & RTF_REJECT) {
3068 switch (rt->dst.error) {
3070 rtm->rtm_type = RTN_BLACKHOLE;
3073 rtm->rtm_type = RTN_PROHIBIT;
3076 rtm->rtm_type = RTN_THROW;
3079 rtm->rtm_type = RTN_UNREACHABLE;
3083 else if (rt->rt6i_flags & RTF_LOCAL)
3084 rtm->rtm_type = RTN_LOCAL;
3085 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3086 rtm->rtm_type = RTN_LOCAL;
3088 rtm->rtm_type = RTN_UNICAST;
3090 if (!netif_carrier_ok(rt->dst.dev)) {
3091 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3092 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3093 rtm->rtm_flags |= RTNH_F_DEAD;
3095 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3096 rtm->rtm_protocol = rt->rt6i_protocol;
3097 if (rt->rt6i_flags & RTF_DYNAMIC)
3098 rtm->rtm_protocol = RTPROT_REDIRECT;
3099 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3100 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3101 rtm->rtm_protocol = RTPROT_RA;
3103 rtm->rtm_protocol = RTPROT_KERNEL;
3106 if (rt->rt6i_flags & RTF_CACHE)
3107 rtm->rtm_flags |= RTM_F_CLONED;
3110 if (nla_put_in6_addr(skb, RTA_DST, dst))
3111 goto nla_put_failure;
3112 rtm->rtm_dst_len = 128;
3113 } else if (rtm->rtm_dst_len)
3114 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3115 goto nla_put_failure;
3116 #ifdef CONFIG_IPV6_SUBTREES
3118 if (nla_put_in6_addr(skb, RTA_SRC, src))
3119 goto nla_put_failure;
3120 rtm->rtm_src_len = 128;
3121 } else if (rtm->rtm_src_len &&
3122 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3123 goto nla_put_failure;
3126 #ifdef CONFIG_IPV6_MROUTE
3127 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3128 int err = ip6mr_get_route(net, skb, rtm, nowait);
3133 goto nla_put_failure;
3135 if (err == -EMSGSIZE)
3136 goto nla_put_failure;
3141 if (nla_put_u32(skb, RTA_IIF, iif))
3142 goto nla_put_failure;
3144 struct in6_addr saddr_buf;
3145 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3146 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3147 goto nla_put_failure;
3150 if (rt->rt6i_prefsrc.plen) {
3151 struct in6_addr saddr_buf;
3152 saddr_buf = rt->rt6i_prefsrc.addr;
3153 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3154 goto nla_put_failure;
3157 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3159 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3160 if (rtnetlink_put_metrics(skb, metrics) < 0)
3161 goto nla_put_failure;
3163 if (rt->rt6i_flags & RTF_GATEWAY) {
3164 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3165 goto nla_put_failure;
3169 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3170 goto nla_put_failure;
3171 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3172 goto nla_put_failure;
3174 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3176 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3177 goto nla_put_failure;
3179 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3180 goto nla_put_failure;
3182 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3184 nlmsg_end(skb, nlh);
3188 nlmsg_cancel(skb, nlh);
3192 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3194 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3197 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3198 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3199 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3203 return rt6_fill_node(arg->net,
3204 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3205 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3206 prefix, 0, NLM_F_MULTI);
3209 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3211 struct net *net = sock_net(in_skb->sk);
3212 struct nlattr *tb[RTA_MAX+1];
3213 struct rt6_info *rt;
3214 struct sk_buff *skb;
3217 int err, iif = 0, oif = 0;
3219 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3224 memset(&fl6, 0, sizeof(fl6));
3227 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3230 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3234 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3237 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3241 iif = nla_get_u32(tb[RTA_IIF]);
3244 oif = nla_get_u32(tb[RTA_OIF]);
3247 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3250 struct net_device *dev;
3253 dev = __dev_get_by_index(net, iif);
3259 fl6.flowi6_iif = iif;
3261 if (!ipv6_addr_any(&fl6.saddr))
3262 flags |= RT6_LOOKUP_F_HAS_SADDR;
3264 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3267 fl6.flowi6_oif = oif;
3269 if (netif_index_is_l3_master(net, oif)) {
3270 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3271 FLOWI_FLAG_SKIP_NH_OIF;
3274 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3277 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3284 /* Reserve room for dummy headers, this skb can pass
3285 through good chunk of routing engine.
3287 skb_reset_mac_header(skb);
3288 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3290 skb_dst_set(skb, &rt->dst);
3292 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3293 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3294 nlh->nlmsg_seq, 0, 0, 0);
3300 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3305 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3306 unsigned int nlm_flags)
3308 struct sk_buff *skb;
3309 struct net *net = info->nl_net;
3314 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3316 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3320 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3321 event, info->portid, seq, 0, 0, nlm_flags);
3323 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3324 WARN_ON(err == -EMSGSIZE);
3328 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3329 info->nlh, gfp_any());
3333 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3336 static int ip6_route_dev_notify(struct notifier_block *this,
3337 unsigned long event, void *ptr)
3339 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3340 struct net *net = dev_net(dev);
3342 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3343 net->ipv6.ip6_null_entry->dst.dev = dev;
3344 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3345 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3346 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3347 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3348 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3349 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3360 #ifdef CONFIG_PROC_FS
3362 static const struct file_operations ipv6_route_proc_fops = {
3363 .owner = THIS_MODULE,
3364 .open = ipv6_route_open,
3366 .llseek = seq_lseek,
3367 .release = seq_release_net,
3370 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3372 struct net *net = (struct net *)seq->private;
3373 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3374 net->ipv6.rt6_stats->fib_nodes,
3375 net->ipv6.rt6_stats->fib_route_nodes,
3376 net->ipv6.rt6_stats->fib_rt_alloc,
3377 net->ipv6.rt6_stats->fib_rt_entries,
3378 net->ipv6.rt6_stats->fib_rt_cache,
3379 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3380 net->ipv6.rt6_stats->fib_discarded_routes);
3385 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3387 return single_open_net(inode, file, rt6_stats_seq_show);
3390 static const struct file_operations rt6_stats_seq_fops = {
3391 .owner = THIS_MODULE,
3392 .open = rt6_stats_seq_open,
3394 .llseek = seq_lseek,
3395 .release = single_release_net,
3397 #endif /* CONFIG_PROC_FS */
3399 #ifdef CONFIG_SYSCTL
3402 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3403 void __user *buffer, size_t *lenp, loff_t *ppos)
3410 net = (struct net *)ctl->extra1;
3411 delay = net->ipv6.sysctl.flush_delay;
3412 proc_dointvec(ctl, write, buffer, lenp, ppos);
3413 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3417 struct ctl_table ipv6_route_table_template[] = {
3419 .procname = "flush",
3420 .data = &init_net.ipv6.sysctl.flush_delay,
3421 .maxlen = sizeof(int),
3423 .proc_handler = ipv6_sysctl_rtcache_flush
3426 .procname = "gc_thresh",
3427 .data = &ip6_dst_ops_template.gc_thresh,
3428 .maxlen = sizeof(int),
3430 .proc_handler = proc_dointvec,
3433 .procname = "max_size",
3434 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3435 .maxlen = sizeof(int),
3437 .proc_handler = proc_dointvec,
3440 .procname = "gc_min_interval",
3441 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3442 .maxlen = sizeof(int),
3444 .proc_handler = proc_dointvec_jiffies,
3447 .procname = "gc_timeout",
3448 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3449 .maxlen = sizeof(int),
3451 .proc_handler = proc_dointvec_jiffies,
3454 .procname = "gc_interval",
3455 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3456 .maxlen = sizeof(int),
3458 .proc_handler = proc_dointvec_jiffies,
3461 .procname = "gc_elasticity",
3462 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3463 .maxlen = sizeof(int),
3465 .proc_handler = proc_dointvec,
3468 .procname = "mtu_expires",
3469 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3470 .maxlen = sizeof(int),
3472 .proc_handler = proc_dointvec_jiffies,
3475 .procname = "min_adv_mss",
3476 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3477 .maxlen = sizeof(int),
3479 .proc_handler = proc_dointvec,
3482 .procname = "gc_min_interval_ms",
3483 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3484 .maxlen = sizeof(int),
3486 .proc_handler = proc_dointvec_ms_jiffies,
3491 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3493 struct ctl_table *table;
3495 table = kmemdup(ipv6_route_table_template,
3496 sizeof(ipv6_route_table_template),
3500 table[0].data = &net->ipv6.sysctl.flush_delay;
3501 table[0].extra1 = net;
3502 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3503 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3504 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3505 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3506 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3507 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3508 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3509 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3510 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3512 /* Don't export sysctls to unprivileged users */
3513 if (net->user_ns != &init_user_ns)
3514 table[0].procname = NULL;
3521 static int __net_init ip6_route_net_init(struct net *net)
3525 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3526 sizeof(net->ipv6.ip6_dst_ops));
3528 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3529 goto out_ip6_dst_ops;
3531 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3532 sizeof(*net->ipv6.ip6_null_entry),
3534 if (!net->ipv6.ip6_null_entry)
3535 goto out_ip6_dst_entries;
3536 net->ipv6.ip6_null_entry->dst.path =
3537 (struct dst_entry *)net->ipv6.ip6_null_entry;
3538 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3539 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3540 ip6_template_metrics, true);
3542 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3543 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3544 sizeof(*net->ipv6.ip6_prohibit_entry),
3546 if (!net->ipv6.ip6_prohibit_entry)
3547 goto out_ip6_null_entry;
3548 net->ipv6.ip6_prohibit_entry->dst.path =
3549 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3550 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3551 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3552 ip6_template_metrics, true);
3554 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3555 sizeof(*net->ipv6.ip6_blk_hole_entry),
3557 if (!net->ipv6.ip6_blk_hole_entry)
3558 goto out_ip6_prohibit_entry;
3559 net->ipv6.ip6_blk_hole_entry->dst.path =
3560 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3561 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3562 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3563 ip6_template_metrics, true);
3566 net->ipv6.sysctl.flush_delay = 0;
3567 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3568 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3569 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3570 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3571 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3572 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3573 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3575 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3581 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3582 out_ip6_prohibit_entry:
3583 kfree(net->ipv6.ip6_prohibit_entry);
3585 kfree(net->ipv6.ip6_null_entry);
3587 out_ip6_dst_entries:
3588 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3593 static void __net_exit ip6_route_net_exit(struct net *net)
3595 kfree(net->ipv6.ip6_null_entry);
3596 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3597 kfree(net->ipv6.ip6_prohibit_entry);
3598 kfree(net->ipv6.ip6_blk_hole_entry);
3600 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3603 static int __net_init ip6_route_net_init_late(struct net *net)
3605 #ifdef CONFIG_PROC_FS
3606 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3607 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3612 static void __net_exit ip6_route_net_exit_late(struct net *net)
3614 #ifdef CONFIG_PROC_FS
3615 remove_proc_entry("ipv6_route", net->proc_net);
3616 remove_proc_entry("rt6_stats", net->proc_net);
3620 static struct pernet_operations ip6_route_net_ops = {
3621 .init = ip6_route_net_init,
3622 .exit = ip6_route_net_exit,
3625 static int __net_init ipv6_inetpeer_init(struct net *net)
3627 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3631 inet_peer_base_init(bp);
3632 net->ipv6.peers = bp;
3636 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3638 struct inet_peer_base *bp = net->ipv6.peers;
3640 net->ipv6.peers = NULL;
3641 inetpeer_invalidate_tree(bp);
3645 static struct pernet_operations ipv6_inetpeer_ops = {
3646 .init = ipv6_inetpeer_init,
3647 .exit = ipv6_inetpeer_exit,
3650 static struct pernet_operations ip6_route_net_late_ops = {
3651 .init = ip6_route_net_init_late,
3652 .exit = ip6_route_net_exit_late,
3655 static struct notifier_block ip6_route_dev_notifier = {
3656 .notifier_call = ip6_route_dev_notify,
3660 int __init ip6_route_init(void)
3666 ip6_dst_ops_template.kmem_cachep =
3667 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3668 SLAB_HWCACHE_ALIGN, NULL);
3669 if (!ip6_dst_ops_template.kmem_cachep)
3672 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3674 goto out_kmem_cache;
3676 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3678 goto out_dst_entries;
3680 ret = register_pernet_subsys(&ip6_route_net_ops);
3682 goto out_register_inetpeer;
3684 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3686 /* Registering of the loopback is done before this portion of code,
3687 * the loopback reference in rt6_info will not be taken, do it
3688 * manually for init_net */
3689 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3690 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3691 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3692 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3693 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3694 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3695 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3699 goto out_register_subsys;
3705 ret = fib6_rules_init();
3709 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3711 goto fib6_rules_init;
3714 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3715 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3716 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3717 goto out_register_late_subsys;
3719 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3721 goto out_register_late_subsys;
3723 for_each_possible_cpu(cpu) {
3724 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3726 INIT_LIST_HEAD(&ul->head);
3727 spin_lock_init(&ul->lock);
3733 out_register_late_subsys:
3734 unregister_pernet_subsys(&ip6_route_net_late_ops);
3736 fib6_rules_cleanup();
3741 out_register_subsys:
3742 unregister_pernet_subsys(&ip6_route_net_ops);
3743 out_register_inetpeer:
3744 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3746 dst_entries_destroy(&ip6_dst_blackhole_ops);
3748 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3752 void ip6_route_cleanup(void)
3754 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3755 unregister_pernet_subsys(&ip6_route_net_late_ops);
3756 fib6_rules_cleanup();
3759 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3760 unregister_pernet_subsys(&ip6_route_net_ops);
3761 dst_entries_destroy(&ip6_dst_blackhole_ops);
3762 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);