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>
65 #include <asm/uaccess.h>
68 #include <linux/sysctl.h>
72 RT6_NUD_FAIL_HARD = -3,
73 RT6_NUD_FAIL_PROBE = -2,
74 RT6_NUD_FAIL_DO_RR = -1,
78 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
79 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
80 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
81 static unsigned int ip6_mtu(const struct dst_entry *dst);
82 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
83 static void ip6_dst_destroy(struct dst_entry *);
84 static void ip6_dst_ifdown(struct dst_entry *,
85 struct net_device *dev, int how);
86 static int ip6_dst_gc(struct dst_ops *ops);
88 static int ip6_pkt_discard(struct sk_buff *skb);
89 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb);
90 static int ip6_pkt_prohibit(struct sk_buff *skb);
91 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb);
92 static void ip6_link_failure(struct sk_buff *skb);
93 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
94 struct sk_buff *skb, u32 mtu);
95 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
97 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
98 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
100 #ifdef CONFIG_IPV6_ROUTE_INFO
101 static struct rt6_info *rt6_add_route_info(struct net *net,
102 const struct in6_addr *prefix, int prefixlen,
103 const struct in6_addr *gwaddr, int ifindex,
105 static struct rt6_info *rt6_get_route_info(struct net *net,
106 const struct in6_addr *prefix, int prefixlen,
107 const struct in6_addr *gwaddr, int ifindex);
110 struct uncached_list {
112 struct list_head head;
115 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
117 static void rt6_uncached_list_add(struct rt6_info *rt)
119 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
121 rt->dst.flags |= DST_NOCACHE;
122 rt->rt6i_uncached_list = ul;
124 spin_lock_bh(&ul->lock);
125 list_add_tail(&rt->rt6i_uncached, &ul->head);
126 spin_unlock_bh(&ul->lock);
129 static void rt6_uncached_list_del(struct rt6_info *rt)
131 if (!list_empty(&rt->rt6i_uncached)) {
132 struct uncached_list *ul = rt->rt6i_uncached_list;
134 spin_lock_bh(&ul->lock);
135 list_del(&rt->rt6i_uncached);
136 spin_unlock_bh(&ul->lock);
140 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
142 struct net_device *loopback_dev = net->loopback_dev;
145 for_each_possible_cpu(cpu) {
146 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
149 spin_lock_bh(&ul->lock);
150 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
151 struct inet6_dev *rt_idev = rt->rt6i_idev;
152 struct net_device *rt_dev = rt->dst.dev;
154 if (rt_idev && (rt_idev->dev == dev || !dev) &&
155 rt_idev->dev != loopback_dev) {
156 rt->rt6i_idev = in6_dev_get(loopback_dev);
157 in6_dev_put(rt_idev);
160 if (rt_dev && (rt_dev == dev || !dev) &&
161 rt_dev != loopback_dev) {
162 rt->dst.dev = loopback_dev;
163 dev_hold(rt->dst.dev);
167 spin_unlock_bh(&ul->lock);
171 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
173 return dst_metrics_write_ptr(rt->dst.from);
176 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
178 struct rt6_info *rt = (struct rt6_info *)dst;
180 if (rt->rt6i_flags & RTF_PCPU)
181 return rt6_pcpu_cow_metrics(rt);
182 else if (rt->rt6i_flags & RTF_CACHE)
185 return dst_cow_metrics_generic(dst, old);
188 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
192 struct in6_addr *p = &rt->rt6i_gateway;
194 if (!ipv6_addr_any(p))
195 return (const void *) p;
197 return &ipv6_hdr(skb)->daddr;
201 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
205 struct rt6_info *rt = (struct rt6_info *) dst;
208 daddr = choose_neigh_daddr(rt, skb, daddr);
209 n = __ipv6_neigh_lookup(dst->dev, daddr);
212 return neigh_create(&nd_tbl, daddr, dst->dev);
215 static struct dst_ops ip6_dst_ops_template = {
219 .check = ip6_dst_check,
220 .default_advmss = ip6_default_advmss,
222 .cow_metrics = ipv6_cow_metrics,
223 .destroy = ip6_dst_destroy,
224 .ifdown = ip6_dst_ifdown,
225 .negative_advice = ip6_negative_advice,
226 .link_failure = ip6_link_failure,
227 .update_pmtu = ip6_rt_update_pmtu,
228 .redirect = rt6_do_redirect,
229 .local_out = __ip6_local_out,
230 .neigh_lookup = ip6_neigh_lookup,
233 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
235 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
237 return mtu ? : dst->dev->mtu;
240 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
241 struct sk_buff *skb, u32 mtu)
245 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
250 static u32 *ip6_rt_blackhole_cow_metrics(struct dst_entry *dst,
256 static struct dst_ops ip6_dst_blackhole_ops = {
258 .destroy = ip6_dst_destroy,
259 .check = ip6_dst_check,
260 .mtu = ip6_blackhole_mtu,
261 .default_advmss = ip6_default_advmss,
262 .update_pmtu = ip6_rt_blackhole_update_pmtu,
263 .redirect = ip6_rt_blackhole_redirect,
264 .cow_metrics = ip6_rt_blackhole_cow_metrics,
265 .neigh_lookup = ip6_neigh_lookup,
268 static const u32 ip6_template_metrics[RTAX_MAX] = {
269 [RTAX_HOPLIMIT - 1] = 0,
272 static const struct rt6_info ip6_null_entry_template = {
274 .__refcnt = ATOMIC_INIT(1),
276 .obsolete = DST_OBSOLETE_FORCE_CHK,
277 .error = -ENETUNREACH,
278 .input = ip6_pkt_discard,
279 .output = ip6_pkt_discard_out,
281 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
282 .rt6i_protocol = RTPROT_KERNEL,
283 .rt6i_metric = ~(u32) 0,
284 .rt6i_ref = ATOMIC_INIT(1),
287 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
289 static const struct rt6_info ip6_prohibit_entry_template = {
291 .__refcnt = ATOMIC_INIT(1),
293 .obsolete = DST_OBSOLETE_FORCE_CHK,
295 .input = ip6_pkt_prohibit,
296 .output = ip6_pkt_prohibit_out,
298 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
299 .rt6i_protocol = RTPROT_KERNEL,
300 .rt6i_metric = ~(u32) 0,
301 .rt6i_ref = ATOMIC_INIT(1),
304 static const struct rt6_info ip6_blk_hole_entry_template = {
306 .__refcnt = ATOMIC_INIT(1),
308 .obsolete = DST_OBSOLETE_FORCE_CHK,
310 .input = dst_discard,
311 .output = dst_discard_sk,
313 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
314 .rt6i_protocol = RTPROT_KERNEL,
315 .rt6i_metric = ~(u32) 0,
316 .rt6i_ref = ATOMIC_INIT(1),
321 /* allocate dst with ip6_dst_ops */
322 static struct rt6_info *__ip6_dst_alloc(struct net *net,
323 struct net_device *dev,
326 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
327 0, DST_OBSOLETE_FORCE_CHK, flags);
330 struct dst_entry *dst = &rt->dst;
332 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
333 INIT_LIST_HEAD(&rt->rt6i_siblings);
334 INIT_LIST_HEAD(&rt->rt6i_uncached);
339 static struct rt6_info *ip6_dst_alloc(struct net *net,
340 struct net_device *dev,
343 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
346 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
350 for_each_possible_cpu(cpu) {
353 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
354 /* no one shares rt */
358 dst_destroy((struct dst_entry *)rt);
366 static void ip6_dst_destroy(struct dst_entry *dst)
368 struct rt6_info *rt = (struct rt6_info *)dst;
369 struct dst_entry *from = dst->from;
370 struct inet6_dev *idev;
372 dst_destroy_metrics_generic(dst);
373 free_percpu(rt->rt6i_pcpu);
374 rt6_uncached_list_del(rt);
376 idev = rt->rt6i_idev;
378 rt->rt6i_idev = NULL;
386 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
389 struct rt6_info *rt = (struct rt6_info *)dst;
390 struct inet6_dev *idev = rt->rt6i_idev;
391 struct net_device *loopback_dev =
392 dev_net(dev)->loopback_dev;
394 if (dev != loopback_dev) {
395 if (idev && idev->dev == dev) {
396 struct inet6_dev *loopback_idev =
397 in6_dev_get(loopback_dev);
399 rt->rt6i_idev = loopback_idev;
406 static bool rt6_check_expired(const struct rt6_info *rt)
408 if (rt->rt6i_flags & RTF_EXPIRES) {
409 if (time_after(jiffies, rt->dst.expires))
411 } else if (rt->dst.from) {
412 return rt6_check_expired((struct rt6_info *) rt->dst.from);
417 /* Multipath route selection:
418 * Hash based function using packet header and flowlabel.
419 * Adapted from fib_info_hashfn()
421 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
422 const struct flowi6 *fl6)
424 return get_hash_from_flowi6(fl6) % candidate_count;
427 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
428 struct flowi6 *fl6, int oif,
431 struct rt6_info *sibling, *next_sibling;
434 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
435 /* Don't change the route, if route_choosen == 0
436 * (siblings does not include ourself)
439 list_for_each_entry_safe(sibling, next_sibling,
440 &match->rt6i_siblings, rt6i_siblings) {
442 if (route_choosen == 0) {
443 if (rt6_score_route(sibling, oif, strict) < 0)
453 * Route lookup. Any table->tb6_lock is implied.
456 static inline struct rt6_info *rt6_device_match(struct net *net,
458 const struct in6_addr *saddr,
462 struct rt6_info *local = NULL;
463 struct rt6_info *sprt;
465 if (!oif && ipv6_addr_any(saddr))
468 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
469 struct net_device *dev = sprt->dst.dev;
472 if (dev->ifindex == oif)
474 if (dev->flags & IFF_LOOPBACK) {
475 if (!sprt->rt6i_idev ||
476 sprt->rt6i_idev->dev->ifindex != oif) {
477 if (flags & RT6_LOOKUP_F_IFACE && oif)
479 if (local && (!oif ||
480 local->rt6i_idev->dev->ifindex == oif))
486 if (ipv6_chk_addr(net, saddr, dev,
487 flags & RT6_LOOKUP_F_IFACE))
496 if (flags & RT6_LOOKUP_F_IFACE)
497 return net->ipv6.ip6_null_entry;
503 #ifdef CONFIG_IPV6_ROUTER_PREF
504 struct __rt6_probe_work {
505 struct work_struct work;
506 struct in6_addr target;
507 struct net_device *dev;
510 static void rt6_probe_deferred(struct work_struct *w)
512 struct in6_addr mcaddr;
513 struct __rt6_probe_work *work =
514 container_of(w, struct __rt6_probe_work, work);
516 addrconf_addr_solict_mult(&work->target, &mcaddr);
517 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, NULL);
522 static void rt6_probe(struct rt6_info *rt)
524 struct __rt6_probe_work *work;
525 struct neighbour *neigh;
527 * Okay, this does not seem to be appropriate
528 * for now, however, we need to check if it
529 * is really so; aka Router Reachability Probing.
531 * Router Reachability Probe MUST be rate-limited
532 * to no more than one per minute.
534 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
537 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
539 if (neigh->nud_state & NUD_VALID)
543 write_lock(&neigh->lock);
544 if (!(neigh->nud_state & NUD_VALID) &&
547 rt->rt6i_idev->cnf.rtr_probe_interval)) {
548 work = kmalloc(sizeof(*work), GFP_ATOMIC);
550 __neigh_set_probe_once(neigh);
552 write_unlock(&neigh->lock);
554 work = kmalloc(sizeof(*work), GFP_ATOMIC);
558 INIT_WORK(&work->work, rt6_probe_deferred);
559 work->target = rt->rt6i_gateway;
560 dev_hold(rt->dst.dev);
561 work->dev = rt->dst.dev;
562 schedule_work(&work->work);
566 rcu_read_unlock_bh();
569 static inline void rt6_probe(struct rt6_info *rt)
575 * Default Router Selection (RFC 2461 6.3.6)
577 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
579 struct net_device *dev = rt->dst.dev;
580 if (!oif || dev->ifindex == oif)
582 if ((dev->flags & IFF_LOOPBACK) &&
583 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
588 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
590 struct neighbour *neigh;
591 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
593 if (rt->rt6i_flags & RTF_NONEXTHOP ||
594 !(rt->rt6i_flags & RTF_GATEWAY))
595 return RT6_NUD_SUCCEED;
598 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
600 read_lock(&neigh->lock);
601 if (neigh->nud_state & NUD_VALID)
602 ret = RT6_NUD_SUCCEED;
603 #ifdef CONFIG_IPV6_ROUTER_PREF
604 else if (!(neigh->nud_state & NUD_FAILED))
605 ret = RT6_NUD_SUCCEED;
607 ret = RT6_NUD_FAIL_PROBE;
609 read_unlock(&neigh->lock);
611 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
612 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
614 rcu_read_unlock_bh();
619 static int rt6_score_route(struct rt6_info *rt, int oif,
624 m = rt6_check_dev(rt, oif);
625 if (!m && (strict & RT6_LOOKUP_F_IFACE))
626 return RT6_NUD_FAIL_HARD;
627 #ifdef CONFIG_IPV6_ROUTER_PREF
628 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
630 if (strict & RT6_LOOKUP_F_REACHABLE) {
631 int n = rt6_check_neigh(rt);
638 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
639 int *mpri, struct rt6_info *match,
643 bool match_do_rr = false;
644 struct inet6_dev *idev = rt->rt6i_idev;
645 struct net_device *dev = rt->dst.dev;
647 if (dev && !netif_carrier_ok(dev) &&
648 idev->cnf.ignore_routes_with_linkdown)
651 if (rt6_check_expired(rt))
654 m = rt6_score_route(rt, oif, strict);
655 if (m == RT6_NUD_FAIL_DO_RR) {
657 m = 0; /* lowest valid score */
658 } else if (m == RT6_NUD_FAIL_HARD) {
662 if (strict & RT6_LOOKUP_F_REACHABLE)
665 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
667 *do_rr = match_do_rr;
675 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
676 struct rt6_info *rr_head,
677 u32 metric, int oif, int strict,
680 struct rt6_info *rt, *match, *cont;
685 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
686 if (rt->rt6i_metric != metric) {
691 match = find_match(rt, oif, strict, &mpri, match, do_rr);
694 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
695 if (rt->rt6i_metric != metric) {
700 match = find_match(rt, oif, strict, &mpri, match, do_rr);
706 for (rt = cont; rt; rt = rt->dst.rt6_next)
707 match = find_match(rt, oif, strict, &mpri, match, do_rr);
712 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
714 struct rt6_info *match, *rt0;
720 fn->rr_ptr = rt0 = fn->leaf;
722 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
726 struct rt6_info *next = rt0->dst.rt6_next;
728 /* no entries matched; do round-robin */
729 if (!next || next->rt6i_metric != rt0->rt6i_metric)
736 net = dev_net(rt0->dst.dev);
737 return match ? match : net->ipv6.ip6_null_entry;
740 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
742 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
745 #ifdef CONFIG_IPV6_ROUTE_INFO
746 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
747 const struct in6_addr *gwaddr)
749 struct net *net = dev_net(dev);
750 struct route_info *rinfo = (struct route_info *) opt;
751 struct in6_addr prefix_buf, *prefix;
753 unsigned long lifetime;
756 if (len < sizeof(struct route_info)) {
760 /* Sanity check for prefix_len and length */
761 if (rinfo->length > 3) {
763 } else if (rinfo->prefix_len > 128) {
765 } else if (rinfo->prefix_len > 64) {
766 if (rinfo->length < 2) {
769 } else if (rinfo->prefix_len > 0) {
770 if (rinfo->length < 1) {
775 pref = rinfo->route_pref;
776 if (pref == ICMPV6_ROUTER_PREF_INVALID)
779 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
781 if (rinfo->length == 3)
782 prefix = (struct in6_addr *)rinfo->prefix;
784 /* this function is safe */
785 ipv6_addr_prefix(&prefix_buf,
786 (struct in6_addr *)rinfo->prefix,
788 prefix = &prefix_buf;
791 if (rinfo->prefix_len == 0)
792 rt = rt6_get_dflt_router(gwaddr, dev);
794 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
795 gwaddr, dev->ifindex);
797 if (rt && !lifetime) {
803 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
806 rt->rt6i_flags = RTF_ROUTEINFO |
807 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
810 if (!addrconf_finite_timeout(lifetime))
811 rt6_clean_expires(rt);
813 rt6_set_expires(rt, jiffies + HZ * lifetime);
821 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
822 struct in6_addr *saddr)
824 struct fib6_node *pn;
826 if (fn->fn_flags & RTN_TL_ROOT)
829 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
830 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
833 if (fn->fn_flags & RTN_RTINFO)
838 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
839 struct fib6_table *table,
840 struct flowi6 *fl6, int flags)
842 struct fib6_node *fn;
845 read_lock_bh(&table->tb6_lock);
846 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
849 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
850 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
851 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
852 if (rt == net->ipv6.ip6_null_entry) {
853 fn = fib6_backtrack(fn, &fl6->saddr);
857 dst_use(&rt->dst, jiffies);
858 read_unlock_bh(&table->tb6_lock);
863 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
866 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
868 EXPORT_SYMBOL_GPL(ip6_route_lookup);
870 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
871 const struct in6_addr *saddr, int oif, int strict)
873 struct flowi6 fl6 = {
877 struct dst_entry *dst;
878 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
881 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
882 flags |= RT6_LOOKUP_F_HAS_SADDR;
885 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
887 return (struct rt6_info *) dst;
893 EXPORT_SYMBOL(rt6_lookup);
895 /* ip6_ins_rt is called with FREE table->tb6_lock.
896 It takes new route entry, the addition fails by any reason the
897 route is freed. In any case, if caller does not hold it, it may
901 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
902 struct mx6_config *mxc)
905 struct fib6_table *table;
907 table = rt->rt6i_table;
908 write_lock_bh(&table->tb6_lock);
909 err = fib6_add(&table->tb6_root, rt, info, mxc);
910 write_unlock_bh(&table->tb6_lock);
915 int ip6_ins_rt(struct rt6_info *rt)
917 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
918 struct mx6_config mxc = { .mx = NULL, };
920 return __ip6_ins_rt(rt, &info, &mxc);
923 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
924 const struct in6_addr *daddr,
925 const struct in6_addr *saddr)
933 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
934 ort = (struct rt6_info *)ort->dst.from;
936 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
941 ip6_rt_copy_init(rt, ort);
942 rt->rt6i_flags |= RTF_CACHE;
944 rt->dst.flags |= DST_HOST;
945 rt->rt6i_dst.addr = *daddr;
946 rt->rt6i_dst.plen = 128;
948 if (!rt6_is_gw_or_nonexthop(ort)) {
949 if (ort->rt6i_dst.plen != 128 &&
950 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
951 rt->rt6i_flags |= RTF_ANYCAST;
952 #ifdef CONFIG_IPV6_SUBTREES
953 if (rt->rt6i_src.plen && saddr) {
954 rt->rt6i_src.addr = *saddr;
955 rt->rt6i_src.plen = 128;
963 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
965 struct rt6_info *pcpu_rt;
967 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
968 rt->dst.dev, rt->dst.flags);
972 ip6_rt_copy_init(pcpu_rt, rt);
973 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
974 pcpu_rt->rt6i_flags |= RTF_PCPU;
978 /* It should be called with read_lock_bh(&tb6_lock) acquired */
979 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
981 struct rt6_info *pcpu_rt, **p;
983 p = this_cpu_ptr(rt->rt6i_pcpu);
987 dst_hold(&pcpu_rt->dst);
988 rt6_dst_from_metrics_check(pcpu_rt);
993 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
995 struct fib6_table *table = rt->rt6i_table;
996 struct rt6_info *pcpu_rt, *prev, **p;
998 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1000 struct net *net = dev_net(rt->dst.dev);
1002 dst_hold(&net->ipv6.ip6_null_entry->dst);
1003 return net->ipv6.ip6_null_entry;
1006 read_lock_bh(&table->tb6_lock);
1007 if (rt->rt6i_pcpu) {
1008 p = this_cpu_ptr(rt->rt6i_pcpu);
1009 prev = cmpxchg(p, NULL, pcpu_rt);
1011 /* If someone did it before us, return prev instead */
1012 dst_destroy(&pcpu_rt->dst);
1016 /* rt has been removed from the fib6 tree
1017 * before we have a chance to acquire the read_lock.
1018 * In this case, don't brother to create a pcpu rt
1019 * since rt is going away anyway. The next
1020 * dst_check() will trigger a re-lookup.
1022 dst_destroy(&pcpu_rt->dst);
1025 dst_hold(&pcpu_rt->dst);
1026 rt6_dst_from_metrics_check(pcpu_rt);
1027 read_unlock_bh(&table->tb6_lock);
1031 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1032 struct flowi6 *fl6, int flags)
1034 struct fib6_node *fn, *saved_fn;
1035 struct rt6_info *rt;
1038 strict |= flags & RT6_LOOKUP_F_IFACE;
1039 if (net->ipv6.devconf_all->forwarding == 0)
1040 strict |= RT6_LOOKUP_F_REACHABLE;
1042 read_lock_bh(&table->tb6_lock);
1044 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1048 rt = rt6_select(fn, oif, strict);
1049 if (rt->rt6i_nsiblings)
1050 rt = rt6_multipath_select(rt, fl6, oif, strict);
1051 if (rt == net->ipv6.ip6_null_entry) {
1052 fn = fib6_backtrack(fn, &fl6->saddr);
1054 goto redo_rt6_select;
1055 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1056 /* also consider unreachable route */
1057 strict &= ~RT6_LOOKUP_F_REACHABLE;
1059 goto redo_rt6_select;
1064 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1065 dst_use(&rt->dst, jiffies);
1066 read_unlock_bh(&table->tb6_lock);
1068 rt6_dst_from_metrics_check(rt);
1070 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1071 !(rt->rt6i_flags & RTF_GATEWAY))) {
1072 /* Create a RTF_CACHE clone which will not be
1073 * owned by the fib6 tree. It is for the special case where
1074 * the daddr in the skb during the neighbor look-up is different
1075 * from the fl6->daddr used to look-up route here.
1078 struct rt6_info *uncached_rt;
1080 dst_use(&rt->dst, jiffies);
1081 read_unlock_bh(&table->tb6_lock);
1083 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1084 dst_release(&rt->dst);
1087 rt6_uncached_list_add(uncached_rt);
1089 uncached_rt = net->ipv6.ip6_null_entry;
1091 dst_hold(&uncached_rt->dst);
1095 /* Get a percpu copy */
1097 struct rt6_info *pcpu_rt;
1099 rt->dst.lastuse = jiffies;
1101 pcpu_rt = rt6_get_pcpu_route(rt);
1104 read_unlock_bh(&table->tb6_lock);
1106 /* We have to do the read_unlock first
1107 * because rt6_make_pcpu_route() may trigger
1108 * ip6_dst_gc() which will take the write_lock.
1111 read_unlock_bh(&table->tb6_lock);
1112 pcpu_rt = rt6_make_pcpu_route(rt);
1113 dst_release(&rt->dst);
1121 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1122 struct flowi6 *fl6, int flags)
1124 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1127 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1128 struct net_device *dev,
1129 struct flowi6 *fl6, int flags)
1131 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1132 flags |= RT6_LOOKUP_F_IFACE;
1134 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1137 void ip6_route_input(struct sk_buff *skb)
1139 const struct ipv6hdr *iph = ipv6_hdr(skb);
1140 struct net *net = dev_net(skb->dev);
1141 int flags = RT6_LOOKUP_F_HAS_SADDR;
1142 struct ip_tunnel_info *tun_info;
1143 struct flowi6 fl6 = {
1144 .flowi6_iif = skb->dev->ifindex,
1145 .daddr = iph->daddr,
1146 .saddr = iph->saddr,
1147 .flowlabel = ip6_flowinfo(iph),
1148 .flowi6_mark = skb->mark,
1149 .flowi6_proto = iph->nexthdr,
1152 tun_info = skb_tunnel_info(skb);
1153 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1154 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1156 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1159 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1160 struct flowi6 *fl6, int flags)
1162 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1165 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1170 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1172 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr))
1173 flags |= RT6_LOOKUP_F_IFACE;
1175 if (!ipv6_addr_any(&fl6->saddr))
1176 flags |= RT6_LOOKUP_F_HAS_SADDR;
1178 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1180 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1182 EXPORT_SYMBOL(ip6_route_output);
1184 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1186 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1187 struct dst_entry *new = NULL;
1189 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1193 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1196 new->input = dst_discard;
1197 new->output = dst_discard_sk;
1199 if (dst_metrics_read_only(&ort->dst))
1200 new->_metrics = ort->dst._metrics;
1202 dst_copy_metrics(new, &ort->dst);
1203 rt->rt6i_idev = ort->rt6i_idev;
1205 in6_dev_hold(rt->rt6i_idev);
1207 rt->rt6i_gateway = ort->rt6i_gateway;
1208 rt->rt6i_flags = ort->rt6i_flags;
1209 rt->rt6i_metric = 0;
1211 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1212 #ifdef CONFIG_IPV6_SUBTREES
1213 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1219 dst_release(dst_orig);
1220 return new ? new : ERR_PTR(-ENOMEM);
1224 * Destination cache support functions
1227 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1230 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1231 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1234 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1236 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1239 if (rt6_check_expired(rt))
1245 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1247 if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1248 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1254 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1256 struct rt6_info *rt;
1258 rt = (struct rt6_info *) dst;
1260 /* All IPV6 dsts are created with ->obsolete set to the value
1261 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1262 * into this function always.
1265 rt6_dst_from_metrics_check(rt);
1267 if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1268 return rt6_dst_from_check(rt, cookie);
1270 return rt6_check(rt, cookie);
1273 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1275 struct rt6_info *rt = (struct rt6_info *) dst;
1278 if (rt->rt6i_flags & RTF_CACHE) {
1279 if (rt6_check_expired(rt)) {
1291 static void ip6_link_failure(struct sk_buff *skb)
1293 struct rt6_info *rt;
1295 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1297 rt = (struct rt6_info *) skb_dst(skb);
1299 if (rt->rt6i_flags & RTF_CACHE) {
1302 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1303 rt->rt6i_node->fn_sernum = -1;
1308 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1310 struct net *net = dev_net(rt->dst.dev);
1312 rt->rt6i_flags |= RTF_MODIFIED;
1313 rt->rt6i_pmtu = mtu;
1314 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1317 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1318 const struct ipv6hdr *iph, u32 mtu)
1320 struct rt6_info *rt6 = (struct rt6_info *)dst;
1322 if (rt6->rt6i_flags & RTF_LOCAL)
1326 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1327 if (mtu >= dst_mtu(dst))
1330 if (rt6->rt6i_flags & RTF_CACHE) {
1331 rt6_do_update_pmtu(rt6, mtu);
1333 const struct in6_addr *daddr, *saddr;
1334 struct rt6_info *nrt6;
1337 daddr = &iph->daddr;
1338 saddr = &iph->saddr;
1340 daddr = &sk->sk_v6_daddr;
1341 saddr = &inet6_sk(sk)->saddr;
1345 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1347 rt6_do_update_pmtu(nrt6, mtu);
1349 /* ip6_ins_rt(nrt6) will bump the
1350 * rt6->rt6i_node->fn_sernum
1351 * which will fail the next rt6_check() and
1352 * invalidate the sk->sk_dst_cache.
1359 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1360 struct sk_buff *skb, u32 mtu)
1362 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1365 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1368 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1369 struct dst_entry *dst;
1372 memset(&fl6, 0, sizeof(fl6));
1373 fl6.flowi6_oif = oif;
1374 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1375 fl6.daddr = iph->daddr;
1376 fl6.saddr = iph->saddr;
1377 fl6.flowlabel = ip6_flowinfo(iph);
1379 dst = ip6_route_output(net, NULL, &fl6);
1381 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1384 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1386 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1388 ip6_update_pmtu(skb, sock_net(sk), mtu,
1389 sk->sk_bound_dev_if, sk->sk_mark);
1391 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1393 /* Handle redirects */
1394 struct ip6rd_flowi {
1396 struct in6_addr gateway;
1399 static struct rt6_info *__ip6_route_redirect(struct net *net,
1400 struct fib6_table *table,
1404 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1405 struct rt6_info *rt;
1406 struct fib6_node *fn;
1408 /* Get the "current" route for this destination and
1409 * check if the redirect has come from approriate router.
1411 * RFC 4861 specifies that redirects should only be
1412 * accepted if they come from the nexthop to the target.
1413 * Due to the way the routes are chosen, this notion
1414 * is a bit fuzzy and one might need to check all possible
1418 read_lock_bh(&table->tb6_lock);
1419 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1421 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1422 if (rt6_check_expired(rt))
1426 if (!(rt->rt6i_flags & RTF_GATEWAY))
1428 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1430 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1436 rt = net->ipv6.ip6_null_entry;
1437 else if (rt->dst.error) {
1438 rt = net->ipv6.ip6_null_entry;
1442 if (rt == net->ipv6.ip6_null_entry) {
1443 fn = fib6_backtrack(fn, &fl6->saddr);
1451 read_unlock_bh(&table->tb6_lock);
1456 static struct dst_entry *ip6_route_redirect(struct net *net,
1457 const struct flowi6 *fl6,
1458 const struct in6_addr *gateway)
1460 int flags = RT6_LOOKUP_F_HAS_SADDR;
1461 struct ip6rd_flowi rdfl;
1464 rdfl.gateway = *gateway;
1466 return fib6_rule_lookup(net, &rdfl.fl6,
1467 flags, __ip6_route_redirect);
1470 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1472 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1473 struct dst_entry *dst;
1476 memset(&fl6, 0, sizeof(fl6));
1477 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1478 fl6.flowi6_oif = oif;
1479 fl6.flowi6_mark = mark;
1480 fl6.daddr = iph->daddr;
1481 fl6.saddr = iph->saddr;
1482 fl6.flowlabel = ip6_flowinfo(iph);
1484 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1485 rt6_do_redirect(dst, NULL, skb);
1488 EXPORT_SYMBOL_GPL(ip6_redirect);
1490 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1493 const struct ipv6hdr *iph = ipv6_hdr(skb);
1494 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1495 struct dst_entry *dst;
1498 memset(&fl6, 0, sizeof(fl6));
1499 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1500 fl6.flowi6_oif = oif;
1501 fl6.flowi6_mark = mark;
1502 fl6.daddr = msg->dest;
1503 fl6.saddr = iph->daddr;
1505 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1506 rt6_do_redirect(dst, NULL, skb);
1510 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1512 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1514 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1516 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1518 struct net_device *dev = dst->dev;
1519 unsigned int mtu = dst_mtu(dst);
1520 struct net *net = dev_net(dev);
1522 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1524 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1525 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1528 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1529 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1530 * IPV6_MAXPLEN is also valid and means: "any MSS,
1531 * rely only on pmtu discovery"
1533 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1538 static unsigned int ip6_mtu(const struct dst_entry *dst)
1540 const struct rt6_info *rt = (const struct rt6_info *)dst;
1541 unsigned int mtu = rt->rt6i_pmtu;
1542 struct inet6_dev *idev;
1547 mtu = dst_metric_raw(dst, RTAX_MTU);
1554 idev = __in6_dev_get(dst->dev);
1556 mtu = idev->cnf.mtu6;
1560 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1563 static struct dst_entry *icmp6_dst_gc_list;
1564 static DEFINE_SPINLOCK(icmp6_dst_lock);
1566 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1569 struct dst_entry *dst;
1570 struct rt6_info *rt;
1571 struct inet6_dev *idev = in6_dev_get(dev);
1572 struct net *net = dev_net(dev);
1574 if (unlikely(!idev))
1575 return ERR_PTR(-ENODEV);
1577 rt = ip6_dst_alloc(net, dev, 0);
1578 if (unlikely(!rt)) {
1580 dst = ERR_PTR(-ENOMEM);
1584 rt->dst.flags |= DST_HOST;
1585 rt->dst.output = ip6_output;
1586 atomic_set(&rt->dst.__refcnt, 1);
1587 rt->rt6i_gateway = fl6->daddr;
1588 rt->rt6i_dst.addr = fl6->daddr;
1589 rt->rt6i_dst.plen = 128;
1590 rt->rt6i_idev = idev;
1591 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1593 spin_lock_bh(&icmp6_dst_lock);
1594 rt->dst.next = icmp6_dst_gc_list;
1595 icmp6_dst_gc_list = &rt->dst;
1596 spin_unlock_bh(&icmp6_dst_lock);
1598 fib6_force_start_gc(net);
1600 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1606 int icmp6_dst_gc(void)
1608 struct dst_entry *dst, **pprev;
1611 spin_lock_bh(&icmp6_dst_lock);
1612 pprev = &icmp6_dst_gc_list;
1614 while ((dst = *pprev) != NULL) {
1615 if (!atomic_read(&dst->__refcnt)) {
1624 spin_unlock_bh(&icmp6_dst_lock);
1629 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1632 struct dst_entry *dst, **pprev;
1634 spin_lock_bh(&icmp6_dst_lock);
1635 pprev = &icmp6_dst_gc_list;
1636 while ((dst = *pprev) != NULL) {
1637 struct rt6_info *rt = (struct rt6_info *) dst;
1638 if (func(rt, arg)) {
1645 spin_unlock_bh(&icmp6_dst_lock);
1648 static int ip6_dst_gc(struct dst_ops *ops)
1650 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1651 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1652 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1653 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1654 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1655 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1658 entries = dst_entries_get_fast(ops);
1659 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1660 entries <= rt_max_size)
1663 net->ipv6.ip6_rt_gc_expire++;
1664 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1665 entries = dst_entries_get_slow(ops);
1666 if (entries < ops->gc_thresh)
1667 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1669 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1670 return entries > rt_max_size;
1673 static int ip6_convert_metrics(struct mx6_config *mxc,
1674 const struct fib6_config *cfg)
1676 bool ecn_ca = false;
1684 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1688 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1689 int type = nla_type(nla);
1694 if (unlikely(type > RTAX_MAX))
1697 if (type == RTAX_CC_ALGO) {
1698 char tmp[TCP_CA_NAME_MAX];
1700 nla_strlcpy(tmp, nla, sizeof(tmp));
1701 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1702 if (val == TCP_CA_UNSPEC)
1705 val = nla_get_u32(nla);
1707 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1711 __set_bit(type - 1, mxc->mx_valid);
1715 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1716 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1726 int ip6_route_info_create(struct fib6_config *cfg, struct rt6_info **rt_ret)
1729 struct net *net = cfg->fc_nlinfo.nl_net;
1730 struct rt6_info *rt = NULL;
1731 struct net_device *dev = NULL;
1732 struct inet6_dev *idev = NULL;
1733 struct fib6_table *table;
1736 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1738 #ifndef CONFIG_IPV6_SUBTREES
1739 if (cfg->fc_src_len)
1742 if (cfg->fc_ifindex) {
1744 dev = dev_get_by_index(net, cfg->fc_ifindex);
1747 idev = in6_dev_get(dev);
1752 if (cfg->fc_metric == 0)
1753 cfg->fc_metric = IP6_RT_PRIO_USER;
1756 if (cfg->fc_nlinfo.nlh &&
1757 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1758 table = fib6_get_table(net, cfg->fc_table);
1760 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1761 table = fib6_new_table(net, cfg->fc_table);
1764 table = fib6_new_table(net, cfg->fc_table);
1770 rt = ip6_dst_alloc(net, NULL,
1771 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1778 if (cfg->fc_flags & RTF_EXPIRES)
1779 rt6_set_expires(rt, jiffies +
1780 clock_t_to_jiffies(cfg->fc_expires));
1782 rt6_clean_expires(rt);
1784 if (cfg->fc_protocol == RTPROT_UNSPEC)
1785 cfg->fc_protocol = RTPROT_BOOT;
1786 rt->rt6i_protocol = cfg->fc_protocol;
1788 addr_type = ipv6_addr_type(&cfg->fc_dst);
1790 if (addr_type & IPV6_ADDR_MULTICAST)
1791 rt->dst.input = ip6_mc_input;
1792 else if (cfg->fc_flags & RTF_LOCAL)
1793 rt->dst.input = ip6_input;
1795 rt->dst.input = ip6_forward;
1797 rt->dst.output = ip6_output;
1799 if (cfg->fc_encap) {
1800 struct lwtunnel_state *lwtstate;
1802 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1803 cfg->fc_encap, AF_INET6, cfg,
1807 rt->dst.lwtstate = lwtstate_get(lwtstate);
1808 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1809 rt->dst.lwtstate->orig_output = rt->dst.output;
1810 rt->dst.output = lwtunnel_output;
1812 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1813 rt->dst.lwtstate->orig_input = rt->dst.input;
1814 rt->dst.input = lwtunnel_input;
1818 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1819 rt->rt6i_dst.plen = cfg->fc_dst_len;
1820 if (rt->rt6i_dst.plen == 128)
1821 rt->dst.flags |= DST_HOST;
1823 #ifdef CONFIG_IPV6_SUBTREES
1824 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1825 rt->rt6i_src.plen = cfg->fc_src_len;
1828 rt->rt6i_metric = cfg->fc_metric;
1830 /* We cannot add true routes via loopback here,
1831 they would result in kernel looping; promote them to reject routes
1833 if ((cfg->fc_flags & RTF_REJECT) ||
1834 (dev && (dev->flags & IFF_LOOPBACK) &&
1835 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1836 !(cfg->fc_flags & RTF_LOCAL))) {
1837 /* hold loopback dev/idev if we haven't done so. */
1838 if (dev != net->loopback_dev) {
1843 dev = net->loopback_dev;
1845 idev = in6_dev_get(dev);
1851 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1852 switch (cfg->fc_type) {
1854 rt->dst.error = -EINVAL;
1855 rt->dst.output = dst_discard_sk;
1856 rt->dst.input = dst_discard;
1859 rt->dst.error = -EACCES;
1860 rt->dst.output = ip6_pkt_prohibit_out;
1861 rt->dst.input = ip6_pkt_prohibit;
1864 case RTN_UNREACHABLE:
1866 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1867 : (cfg->fc_type == RTN_UNREACHABLE)
1868 ? -EHOSTUNREACH : -ENETUNREACH;
1869 rt->dst.output = ip6_pkt_discard_out;
1870 rt->dst.input = ip6_pkt_discard;
1876 if (cfg->fc_flags & RTF_GATEWAY) {
1877 const struct in6_addr *gw_addr;
1880 gw_addr = &cfg->fc_gateway;
1881 gwa_type = ipv6_addr_type(gw_addr);
1883 /* if gw_addr is local we will fail to detect this in case
1884 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1885 * will return already-added prefix route via interface that
1886 * prefix route was assigned to, which might be non-loopback.
1889 if (ipv6_chk_addr_and_flags(net, gw_addr,
1890 gwa_type & IPV6_ADDR_LINKLOCAL ?
1894 rt->rt6i_gateway = *gw_addr;
1896 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1897 struct rt6_info *grt;
1899 /* IPv6 strictly inhibits using not link-local
1900 addresses as nexthop address.
1901 Otherwise, router will not able to send redirects.
1902 It is very good, but in some (rare!) circumstances
1903 (SIT, PtP, NBMA NOARP links) it is handy to allow
1904 some exceptions. --ANK
1906 if (!(gwa_type & IPV6_ADDR_UNICAST))
1909 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1911 err = -EHOSTUNREACH;
1915 if (dev != grt->dst.dev) {
1921 idev = grt->rt6i_idev;
1923 in6_dev_hold(grt->rt6i_idev);
1925 if (!(grt->rt6i_flags & RTF_GATEWAY))
1933 if (!dev || (dev->flags & IFF_LOOPBACK))
1941 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1942 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1946 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1947 rt->rt6i_prefsrc.plen = 128;
1949 rt->rt6i_prefsrc.plen = 0;
1951 rt->rt6i_flags = cfg->fc_flags;
1955 rt->rt6i_idev = idev;
1956 rt->rt6i_table = table;
1958 cfg->fc_nlinfo.nl_net = dev_net(dev);
1976 int ip6_route_add(struct fib6_config *cfg)
1978 struct mx6_config mxc = { .mx = NULL, };
1979 struct rt6_info *rt = NULL;
1982 err = ip6_route_info_create(cfg, &rt);
1986 err = ip6_convert_metrics(&mxc, cfg);
1990 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2002 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2005 struct fib6_table *table;
2006 struct net *net = dev_net(rt->dst.dev);
2008 if (rt == net->ipv6.ip6_null_entry ||
2009 rt->dst.flags & DST_NOCACHE) {
2014 table = rt->rt6i_table;
2015 write_lock_bh(&table->tb6_lock);
2016 err = fib6_del(rt, info);
2017 write_unlock_bh(&table->tb6_lock);
2024 int ip6_del_rt(struct rt6_info *rt)
2026 struct nl_info info = {
2027 .nl_net = dev_net(rt->dst.dev),
2029 return __ip6_del_rt(rt, &info);
2032 static int ip6_route_del(struct fib6_config *cfg)
2034 struct fib6_table *table;
2035 struct fib6_node *fn;
2036 struct rt6_info *rt;
2039 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2043 read_lock_bh(&table->tb6_lock);
2045 fn = fib6_locate(&table->tb6_root,
2046 &cfg->fc_dst, cfg->fc_dst_len,
2047 &cfg->fc_src, cfg->fc_src_len);
2050 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2051 if ((rt->rt6i_flags & RTF_CACHE) &&
2052 !(cfg->fc_flags & RTF_CACHE))
2054 if (cfg->fc_ifindex &&
2056 rt->dst.dev->ifindex != cfg->fc_ifindex))
2058 if (cfg->fc_flags & RTF_GATEWAY &&
2059 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2061 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2064 read_unlock_bh(&table->tb6_lock);
2066 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2069 read_unlock_bh(&table->tb6_lock);
2074 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2076 struct net *net = dev_net(skb->dev);
2077 struct netevent_redirect netevent;
2078 struct rt6_info *rt, *nrt = NULL;
2079 struct ndisc_options ndopts;
2080 struct inet6_dev *in6_dev;
2081 struct neighbour *neigh;
2083 int optlen, on_link;
2086 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2087 optlen -= sizeof(*msg);
2090 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2094 msg = (struct rd_msg *)icmp6_hdr(skb);
2096 if (ipv6_addr_is_multicast(&msg->dest)) {
2097 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2102 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2104 } else if (ipv6_addr_type(&msg->target) !=
2105 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2106 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2110 in6_dev = __in6_dev_get(skb->dev);
2113 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2117 * The IP source address of the Redirect MUST be the same as the current
2118 * first-hop router for the specified ICMP Destination Address.
2121 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2122 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2127 if (ndopts.nd_opts_tgt_lladdr) {
2128 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2131 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2136 rt = (struct rt6_info *) dst;
2137 if (rt == net->ipv6.ip6_null_entry) {
2138 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2142 /* Redirect received -> path was valid.
2143 * Look, redirects are sent only in response to data packets,
2144 * so that this nexthop apparently is reachable. --ANK
2146 dst_confirm(&rt->dst);
2148 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2153 * We have finally decided to accept it.
2156 neigh_update(neigh, lladdr, NUD_STALE,
2157 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2158 NEIGH_UPDATE_F_OVERRIDE|
2159 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2160 NEIGH_UPDATE_F_ISROUTER))
2163 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2167 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2169 nrt->rt6i_flags &= ~RTF_GATEWAY;
2171 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2173 if (ip6_ins_rt(nrt))
2176 netevent.old = &rt->dst;
2177 netevent.new = &nrt->dst;
2178 netevent.daddr = &msg->dest;
2179 netevent.neigh = neigh;
2180 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2182 if (rt->rt6i_flags & RTF_CACHE) {
2183 rt = (struct rt6_info *) dst_clone(&rt->dst);
2188 neigh_release(neigh);
2192 * Misc support functions
2195 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2197 BUG_ON(from->dst.from);
2199 rt->rt6i_flags &= ~RTF_EXPIRES;
2200 dst_hold(&from->dst);
2201 rt->dst.from = &from->dst;
2202 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2205 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2207 rt->dst.input = ort->dst.input;
2208 rt->dst.output = ort->dst.output;
2209 rt->rt6i_dst = ort->rt6i_dst;
2210 rt->dst.error = ort->dst.error;
2211 rt->rt6i_idev = ort->rt6i_idev;
2213 in6_dev_hold(rt->rt6i_idev);
2214 rt->dst.lastuse = jiffies;
2215 rt->rt6i_gateway = ort->rt6i_gateway;
2216 rt->rt6i_flags = ort->rt6i_flags;
2217 rt6_set_from(rt, ort);
2218 rt->rt6i_metric = ort->rt6i_metric;
2219 #ifdef CONFIG_IPV6_SUBTREES
2220 rt->rt6i_src = ort->rt6i_src;
2222 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2223 rt->rt6i_table = ort->rt6i_table;
2224 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2227 #ifdef CONFIG_IPV6_ROUTE_INFO
2228 static struct rt6_info *rt6_get_route_info(struct net *net,
2229 const struct in6_addr *prefix, int prefixlen,
2230 const struct in6_addr *gwaddr, int ifindex)
2232 struct fib6_node *fn;
2233 struct rt6_info *rt = NULL;
2234 struct fib6_table *table;
2236 table = fib6_get_table(net, RT6_TABLE_INFO);
2240 read_lock_bh(&table->tb6_lock);
2241 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2245 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2246 if (rt->dst.dev->ifindex != ifindex)
2248 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2250 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2256 read_unlock_bh(&table->tb6_lock);
2260 static struct rt6_info *rt6_add_route_info(struct net *net,
2261 const struct in6_addr *prefix, int prefixlen,
2262 const struct in6_addr *gwaddr, int ifindex,
2265 struct fib6_config cfg = {
2266 .fc_table = RT6_TABLE_INFO,
2267 .fc_metric = IP6_RT_PRIO_USER,
2268 .fc_ifindex = ifindex,
2269 .fc_dst_len = prefixlen,
2270 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2271 RTF_UP | RTF_PREF(pref),
2272 .fc_nlinfo.portid = 0,
2273 .fc_nlinfo.nlh = NULL,
2274 .fc_nlinfo.nl_net = net,
2277 cfg.fc_dst = *prefix;
2278 cfg.fc_gateway = *gwaddr;
2280 /* We should treat it as a default route if prefix length is 0. */
2282 cfg.fc_flags |= RTF_DEFAULT;
2284 ip6_route_add(&cfg);
2286 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2290 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2292 struct rt6_info *rt;
2293 struct fib6_table *table;
2295 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2299 read_lock_bh(&table->tb6_lock);
2300 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2301 if (dev == rt->dst.dev &&
2302 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2303 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2308 read_unlock_bh(&table->tb6_lock);
2312 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2313 struct net_device *dev,
2316 struct fib6_config cfg = {
2317 .fc_table = RT6_TABLE_DFLT,
2318 .fc_metric = IP6_RT_PRIO_USER,
2319 .fc_ifindex = dev->ifindex,
2320 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2321 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2322 .fc_nlinfo.portid = 0,
2323 .fc_nlinfo.nlh = NULL,
2324 .fc_nlinfo.nl_net = dev_net(dev),
2327 cfg.fc_gateway = *gwaddr;
2329 ip6_route_add(&cfg);
2331 return rt6_get_dflt_router(gwaddr, dev);
2334 void rt6_purge_dflt_routers(struct net *net)
2336 struct rt6_info *rt;
2337 struct fib6_table *table;
2339 /* NOTE: Keep consistent with rt6_get_dflt_router */
2340 table = fib6_get_table(net, RT6_TABLE_DFLT);
2345 read_lock_bh(&table->tb6_lock);
2346 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2347 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2348 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2350 read_unlock_bh(&table->tb6_lock);
2355 read_unlock_bh(&table->tb6_lock);
2358 static void rtmsg_to_fib6_config(struct net *net,
2359 struct in6_rtmsg *rtmsg,
2360 struct fib6_config *cfg)
2362 memset(cfg, 0, sizeof(*cfg));
2364 cfg->fc_table = RT6_TABLE_MAIN;
2365 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2366 cfg->fc_metric = rtmsg->rtmsg_metric;
2367 cfg->fc_expires = rtmsg->rtmsg_info;
2368 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2369 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2370 cfg->fc_flags = rtmsg->rtmsg_flags;
2372 cfg->fc_nlinfo.nl_net = net;
2374 cfg->fc_dst = rtmsg->rtmsg_dst;
2375 cfg->fc_src = rtmsg->rtmsg_src;
2376 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2379 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2381 struct fib6_config cfg;
2382 struct in6_rtmsg rtmsg;
2386 case SIOCADDRT: /* Add a route */
2387 case SIOCDELRT: /* Delete a route */
2388 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2390 err = copy_from_user(&rtmsg, arg,
2391 sizeof(struct in6_rtmsg));
2395 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2400 err = ip6_route_add(&cfg);
2403 err = ip6_route_del(&cfg);
2417 * Drop the packet on the floor
2420 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2423 struct dst_entry *dst = skb_dst(skb);
2424 switch (ipstats_mib_noroutes) {
2425 case IPSTATS_MIB_INNOROUTES:
2426 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2427 if (type == IPV6_ADDR_ANY) {
2428 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2429 IPSTATS_MIB_INADDRERRORS);
2433 case IPSTATS_MIB_OUTNOROUTES:
2434 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2435 ipstats_mib_noroutes);
2438 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2443 static int ip6_pkt_discard(struct sk_buff *skb)
2445 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2448 static int ip6_pkt_discard_out(struct sock *sk, struct sk_buff *skb)
2450 skb->dev = skb_dst(skb)->dev;
2451 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2454 static int ip6_pkt_prohibit(struct sk_buff *skb)
2456 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2459 static int ip6_pkt_prohibit_out(struct sock *sk, struct sk_buff *skb)
2461 skb->dev = skb_dst(skb)->dev;
2462 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2466 * Allocate a dst for local (unicast / anycast) address.
2469 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2470 const struct in6_addr *addr,
2473 struct net *net = dev_net(idev->dev);
2474 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2477 return ERR_PTR(-ENOMEM);
2481 rt->dst.flags |= DST_HOST;
2482 rt->dst.input = ip6_input;
2483 rt->dst.output = ip6_output;
2484 rt->rt6i_idev = idev;
2486 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2488 rt->rt6i_flags |= RTF_ANYCAST;
2490 rt->rt6i_flags |= RTF_LOCAL;
2492 rt->rt6i_gateway = *addr;
2493 rt->rt6i_dst.addr = *addr;
2494 rt->rt6i_dst.plen = 128;
2495 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2496 rt->dst.flags |= DST_NOCACHE;
2498 atomic_set(&rt->dst.__refcnt, 1);
2503 int ip6_route_get_saddr(struct net *net,
2504 struct rt6_info *rt,
2505 const struct in6_addr *daddr,
2507 struct in6_addr *saddr)
2509 struct inet6_dev *idev =
2510 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2512 if (rt && rt->rt6i_prefsrc.plen)
2513 *saddr = rt->rt6i_prefsrc.addr;
2515 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2516 daddr, prefs, saddr);
2520 /* remove deleted ip from prefsrc entries */
2521 struct arg_dev_net_ip {
2522 struct net_device *dev;
2524 struct in6_addr *addr;
2527 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2529 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2530 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2531 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2533 if (((void *)rt->dst.dev == dev || !dev) &&
2534 rt != net->ipv6.ip6_null_entry &&
2535 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2536 /* remove prefsrc entry */
2537 rt->rt6i_prefsrc.plen = 0;
2542 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2544 struct net *net = dev_net(ifp->idev->dev);
2545 struct arg_dev_net_ip adni = {
2546 .dev = ifp->idev->dev,
2550 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2553 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2554 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2556 /* Remove routers and update dst entries when gateway turn into host. */
2557 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2559 struct in6_addr *gateway = (struct in6_addr *)arg;
2561 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2562 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2563 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2569 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2571 fib6_clean_all(net, fib6_clean_tohost, gateway);
2574 struct arg_dev_net {
2575 struct net_device *dev;
2579 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2581 const struct arg_dev_net *adn = arg;
2582 const struct net_device *dev = adn->dev;
2584 if ((rt->dst.dev == dev || !dev) &&
2585 rt != adn->net->ipv6.ip6_null_entry)
2591 void rt6_ifdown(struct net *net, struct net_device *dev)
2593 struct arg_dev_net adn = {
2598 fib6_clean_all(net, fib6_ifdown, &adn);
2599 icmp6_clean_all(fib6_ifdown, &adn);
2600 rt6_uncached_list_flush_dev(net, dev);
2603 struct rt6_mtu_change_arg {
2604 struct net_device *dev;
2608 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2610 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2611 struct inet6_dev *idev;
2613 /* In IPv6 pmtu discovery is not optional,
2614 so that RTAX_MTU lock cannot disable it.
2615 We still use this lock to block changes
2616 caused by addrconf/ndisc.
2619 idev = __in6_dev_get(arg->dev);
2623 /* For administrative MTU increase, there is no way to discover
2624 IPv6 PMTU increase, so PMTU increase should be updated here.
2625 Since RFC 1981 doesn't include administrative MTU increase
2626 update PMTU increase is a MUST. (i.e. jumbo frame)
2629 If new MTU is less than route PMTU, this new MTU will be the
2630 lowest MTU in the path, update the route PMTU to reflect PMTU
2631 decreases; if new MTU is greater than route PMTU, and the
2632 old MTU is the lowest MTU in the path, update the route PMTU
2633 to reflect the increase. In this case if the other nodes' MTU
2634 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2637 if (rt->dst.dev == arg->dev &&
2638 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2639 if (rt->rt6i_flags & RTF_CACHE) {
2640 /* For RTF_CACHE with rt6i_pmtu == 0
2641 * (i.e. a redirected route),
2642 * the metrics of its rt->dst.from has already
2645 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2646 rt->rt6i_pmtu = arg->mtu;
2647 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2648 (dst_mtu(&rt->dst) < arg->mtu &&
2649 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2650 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2656 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2658 struct rt6_mtu_change_arg arg = {
2663 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2666 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2667 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2668 [RTA_OIF] = { .type = NLA_U32 },
2669 [RTA_IIF] = { .type = NLA_U32 },
2670 [RTA_PRIORITY] = { .type = NLA_U32 },
2671 [RTA_METRICS] = { .type = NLA_NESTED },
2672 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2673 [RTA_PREF] = { .type = NLA_U8 },
2674 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2675 [RTA_ENCAP] = { .type = NLA_NESTED },
2678 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2679 struct fib6_config *cfg)
2682 struct nlattr *tb[RTA_MAX+1];
2686 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2691 rtm = nlmsg_data(nlh);
2692 memset(cfg, 0, sizeof(*cfg));
2694 cfg->fc_table = rtm->rtm_table;
2695 cfg->fc_dst_len = rtm->rtm_dst_len;
2696 cfg->fc_src_len = rtm->rtm_src_len;
2697 cfg->fc_flags = RTF_UP;
2698 cfg->fc_protocol = rtm->rtm_protocol;
2699 cfg->fc_type = rtm->rtm_type;
2701 if (rtm->rtm_type == RTN_UNREACHABLE ||
2702 rtm->rtm_type == RTN_BLACKHOLE ||
2703 rtm->rtm_type == RTN_PROHIBIT ||
2704 rtm->rtm_type == RTN_THROW)
2705 cfg->fc_flags |= RTF_REJECT;
2707 if (rtm->rtm_type == RTN_LOCAL)
2708 cfg->fc_flags |= RTF_LOCAL;
2710 if (rtm->rtm_flags & RTM_F_CLONED)
2711 cfg->fc_flags |= RTF_CACHE;
2713 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2714 cfg->fc_nlinfo.nlh = nlh;
2715 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2717 if (tb[RTA_GATEWAY]) {
2718 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2719 cfg->fc_flags |= RTF_GATEWAY;
2723 int plen = (rtm->rtm_dst_len + 7) >> 3;
2725 if (nla_len(tb[RTA_DST]) < plen)
2728 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2732 int plen = (rtm->rtm_src_len + 7) >> 3;
2734 if (nla_len(tb[RTA_SRC]) < plen)
2737 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2740 if (tb[RTA_PREFSRC])
2741 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2744 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2746 if (tb[RTA_PRIORITY])
2747 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2749 if (tb[RTA_METRICS]) {
2750 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2751 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2755 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2757 if (tb[RTA_MULTIPATH]) {
2758 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2759 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2763 pref = nla_get_u8(tb[RTA_PREF]);
2764 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2765 pref != ICMPV6_ROUTER_PREF_HIGH)
2766 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2767 cfg->fc_flags |= RTF_PREF(pref);
2771 cfg->fc_encap = tb[RTA_ENCAP];
2773 if (tb[RTA_ENCAP_TYPE])
2774 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2782 struct rt6_info *rt6_info;
2783 struct fib6_config r_cfg;
2784 struct mx6_config mxc;
2785 struct list_head next;
2788 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2792 list_for_each_entry(nh, rt6_nh_list, next) {
2793 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2794 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2795 nh->r_cfg.fc_ifindex);
2799 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2800 struct rt6_info *rt, struct fib6_config *r_cfg)
2803 struct rt6_info *rtnh;
2806 list_for_each_entry(nh, rt6_nh_list, next) {
2807 /* check if rt6_info already exists */
2808 rtnh = nh->rt6_info;
2810 if (rtnh->dst.dev == rt->dst.dev &&
2811 rtnh->rt6i_idev == rt->rt6i_idev &&
2812 ipv6_addr_equal(&rtnh->rt6i_gateway,
2817 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2821 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2826 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2827 list_add_tail(&nh->next, rt6_nh_list);
2832 static int ip6_route_multipath_add(struct fib6_config *cfg)
2834 struct fib6_config r_cfg;
2835 struct rtnexthop *rtnh;
2836 struct rt6_info *rt;
2837 struct rt6_nh *err_nh;
2838 struct rt6_nh *nh, *nh_safe;
2843 int replace = (cfg->fc_nlinfo.nlh &&
2844 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2845 LIST_HEAD(rt6_nh_list);
2847 remaining = cfg->fc_mp_len;
2848 rtnh = (struct rtnexthop *)cfg->fc_mp;
2850 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2851 * rt6_info structs per nexthop
2853 while (rtnh_ok(rtnh, remaining)) {
2854 memcpy(&r_cfg, cfg, sizeof(*cfg));
2855 if (rtnh->rtnh_ifindex)
2856 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2858 attrlen = rtnh_attrlen(rtnh);
2860 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2862 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2864 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2865 r_cfg.fc_flags |= RTF_GATEWAY;
2867 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2868 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2870 r_cfg.fc_encap_type = nla_get_u16(nla);
2873 err = ip6_route_info_create(&r_cfg, &rt);
2877 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2883 rtnh = rtnh_next(rtnh, &remaining);
2887 list_for_each_entry(nh, &rt6_nh_list, next) {
2888 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2889 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2890 nh->rt6_info = NULL;
2893 ip6_print_replace_route_err(&rt6_nh_list);
2898 /* Because each route is added like a single route we remove
2899 * these flags after the first nexthop: if there is a collision,
2900 * we have already failed to add the first nexthop:
2901 * fib6_add_rt2node() has rejected it; when replacing, old
2902 * nexthops have been replaced by first new, the rest should
2905 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2913 /* Delete routes that were already added */
2914 list_for_each_entry(nh, &rt6_nh_list, next) {
2917 ip6_route_del(&nh->r_cfg);
2921 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2923 dst_free(&nh->rt6_info->dst);
2925 list_del(&nh->next);
2932 static int ip6_route_multipath_del(struct fib6_config *cfg)
2934 struct fib6_config r_cfg;
2935 struct rtnexthop *rtnh;
2938 int err = 1, last_err = 0;
2940 remaining = cfg->fc_mp_len;
2941 rtnh = (struct rtnexthop *)cfg->fc_mp;
2943 /* Parse a Multipath Entry */
2944 while (rtnh_ok(rtnh, remaining)) {
2945 memcpy(&r_cfg, cfg, sizeof(*cfg));
2946 if (rtnh->rtnh_ifindex)
2947 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2949 attrlen = rtnh_attrlen(rtnh);
2951 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2953 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2955 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2956 r_cfg.fc_flags |= RTF_GATEWAY;
2959 err = ip6_route_del(&r_cfg);
2963 rtnh = rtnh_next(rtnh, &remaining);
2969 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2971 struct fib6_config cfg;
2974 err = rtm_to_fib6_config(skb, nlh, &cfg);
2979 return ip6_route_multipath_del(&cfg);
2981 return ip6_route_del(&cfg);
2984 static int inet6_rtm_newroute(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_add(&cfg);
2996 return ip6_route_add(&cfg);
2999 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3001 return NLMSG_ALIGN(sizeof(struct rtmsg))
3002 + nla_total_size(16) /* RTA_SRC */
3003 + nla_total_size(16) /* RTA_DST */
3004 + nla_total_size(16) /* RTA_GATEWAY */
3005 + nla_total_size(16) /* RTA_PREFSRC */
3006 + nla_total_size(4) /* RTA_TABLE */
3007 + nla_total_size(4) /* RTA_IIF */
3008 + nla_total_size(4) /* RTA_OIF */
3009 + nla_total_size(4) /* RTA_PRIORITY */
3010 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3011 + nla_total_size(sizeof(struct rta_cacheinfo))
3012 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3013 + nla_total_size(1) /* RTA_PREF */
3014 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3017 static int rt6_fill_node(struct net *net,
3018 struct sk_buff *skb, struct rt6_info *rt,
3019 struct in6_addr *dst, struct in6_addr *src,
3020 int iif, int type, u32 portid, u32 seq,
3021 int prefix, int nowait, unsigned int flags)
3023 u32 metrics[RTAX_MAX];
3025 struct nlmsghdr *nlh;
3029 if (prefix) { /* user wants prefix routes only */
3030 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3031 /* success since this is not a prefix route */
3036 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3040 rtm = nlmsg_data(nlh);
3041 rtm->rtm_family = AF_INET6;
3042 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3043 rtm->rtm_src_len = rt->rt6i_src.plen;
3046 table = rt->rt6i_table->tb6_id;
3048 table = RT6_TABLE_UNSPEC;
3049 rtm->rtm_table = table;
3050 if (nla_put_u32(skb, RTA_TABLE, table))
3051 goto nla_put_failure;
3052 if (rt->rt6i_flags & RTF_REJECT) {
3053 switch (rt->dst.error) {
3055 rtm->rtm_type = RTN_BLACKHOLE;
3058 rtm->rtm_type = RTN_PROHIBIT;
3061 rtm->rtm_type = RTN_THROW;
3064 rtm->rtm_type = RTN_UNREACHABLE;
3068 else if (rt->rt6i_flags & RTF_LOCAL)
3069 rtm->rtm_type = RTN_LOCAL;
3070 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3071 rtm->rtm_type = RTN_LOCAL;
3073 rtm->rtm_type = RTN_UNICAST;
3075 if (!netif_carrier_ok(rt->dst.dev)) {
3076 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3077 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3078 rtm->rtm_flags |= RTNH_F_DEAD;
3080 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3081 rtm->rtm_protocol = rt->rt6i_protocol;
3082 if (rt->rt6i_flags & RTF_DYNAMIC)
3083 rtm->rtm_protocol = RTPROT_REDIRECT;
3084 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3085 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3086 rtm->rtm_protocol = RTPROT_RA;
3088 rtm->rtm_protocol = RTPROT_KERNEL;
3091 if (rt->rt6i_flags & RTF_CACHE)
3092 rtm->rtm_flags |= RTM_F_CLONED;
3095 if (nla_put_in6_addr(skb, RTA_DST, dst))
3096 goto nla_put_failure;
3097 rtm->rtm_dst_len = 128;
3098 } else if (rtm->rtm_dst_len)
3099 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3100 goto nla_put_failure;
3101 #ifdef CONFIG_IPV6_SUBTREES
3103 if (nla_put_in6_addr(skb, RTA_SRC, src))
3104 goto nla_put_failure;
3105 rtm->rtm_src_len = 128;
3106 } else if (rtm->rtm_src_len &&
3107 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3108 goto nla_put_failure;
3111 #ifdef CONFIG_IPV6_MROUTE
3112 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3113 int err = ip6mr_get_route(net, skb, rtm, nowait);
3118 goto nla_put_failure;
3120 if (err == -EMSGSIZE)
3121 goto nla_put_failure;
3126 if (nla_put_u32(skb, RTA_IIF, iif))
3127 goto nla_put_failure;
3129 struct in6_addr saddr_buf;
3130 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3131 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3132 goto nla_put_failure;
3135 if (rt->rt6i_prefsrc.plen) {
3136 struct in6_addr saddr_buf;
3137 saddr_buf = rt->rt6i_prefsrc.addr;
3138 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3139 goto nla_put_failure;
3142 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3144 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3145 if (rtnetlink_put_metrics(skb, metrics) < 0)
3146 goto nla_put_failure;
3148 if (rt->rt6i_flags & RTF_GATEWAY) {
3149 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3150 goto nla_put_failure;
3154 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3155 goto nla_put_failure;
3156 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3157 goto nla_put_failure;
3159 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3161 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3162 goto nla_put_failure;
3164 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3165 goto nla_put_failure;
3167 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3169 nlmsg_end(skb, nlh);
3173 nlmsg_cancel(skb, nlh);
3177 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3179 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3182 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3183 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3184 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3188 return rt6_fill_node(arg->net,
3189 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3190 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3191 prefix, 0, NLM_F_MULTI);
3194 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3196 struct net *net = sock_net(in_skb->sk);
3197 struct nlattr *tb[RTA_MAX+1];
3198 struct rt6_info *rt;
3199 struct sk_buff *skb;
3202 int err, iif = 0, oif = 0;
3204 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3209 memset(&fl6, 0, sizeof(fl6));
3212 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3215 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3219 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3222 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3226 iif = nla_get_u32(tb[RTA_IIF]);
3229 oif = nla_get_u32(tb[RTA_OIF]);
3232 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3235 struct net_device *dev;
3238 dev = __dev_get_by_index(net, iif);
3244 fl6.flowi6_iif = iif;
3246 if (!ipv6_addr_any(&fl6.saddr))
3247 flags |= RT6_LOOKUP_F_HAS_SADDR;
3249 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3252 fl6.flowi6_oif = oif;
3254 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3257 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3264 /* Reserve room for dummy headers, this skb can pass
3265 through good chunk of routing engine.
3267 skb_reset_mac_header(skb);
3268 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3270 skb_dst_set(skb, &rt->dst);
3272 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3273 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3274 nlh->nlmsg_seq, 0, 0, 0);
3280 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3285 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3286 unsigned int nlm_flags)
3288 struct sk_buff *skb;
3289 struct net *net = info->nl_net;
3294 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3296 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3300 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3301 event, info->portid, seq, 0, 0, nlm_flags);
3303 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3304 WARN_ON(err == -EMSGSIZE);
3308 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3309 info->nlh, gfp_any());
3313 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3316 static int ip6_route_dev_notify(struct notifier_block *this,
3317 unsigned long event, void *ptr)
3319 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3320 struct net *net = dev_net(dev);
3322 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3323 net->ipv6.ip6_null_entry->dst.dev = dev;
3324 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3325 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3326 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3327 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3328 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3329 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3340 #ifdef CONFIG_PROC_FS
3342 static const struct file_operations ipv6_route_proc_fops = {
3343 .owner = THIS_MODULE,
3344 .open = ipv6_route_open,
3346 .llseek = seq_lseek,
3347 .release = seq_release_net,
3350 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3352 struct net *net = (struct net *)seq->private;
3353 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3354 net->ipv6.rt6_stats->fib_nodes,
3355 net->ipv6.rt6_stats->fib_route_nodes,
3356 net->ipv6.rt6_stats->fib_rt_alloc,
3357 net->ipv6.rt6_stats->fib_rt_entries,
3358 net->ipv6.rt6_stats->fib_rt_cache,
3359 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3360 net->ipv6.rt6_stats->fib_discarded_routes);
3365 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3367 return single_open_net(inode, file, rt6_stats_seq_show);
3370 static const struct file_operations rt6_stats_seq_fops = {
3371 .owner = THIS_MODULE,
3372 .open = rt6_stats_seq_open,
3374 .llseek = seq_lseek,
3375 .release = single_release_net,
3377 #endif /* CONFIG_PROC_FS */
3379 #ifdef CONFIG_SYSCTL
3382 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3383 void __user *buffer, size_t *lenp, loff_t *ppos)
3390 net = (struct net *)ctl->extra1;
3391 delay = net->ipv6.sysctl.flush_delay;
3392 proc_dointvec(ctl, write, buffer, lenp, ppos);
3393 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3397 struct ctl_table ipv6_route_table_template[] = {
3399 .procname = "flush",
3400 .data = &init_net.ipv6.sysctl.flush_delay,
3401 .maxlen = sizeof(int),
3403 .proc_handler = ipv6_sysctl_rtcache_flush
3406 .procname = "gc_thresh",
3407 .data = &ip6_dst_ops_template.gc_thresh,
3408 .maxlen = sizeof(int),
3410 .proc_handler = proc_dointvec,
3413 .procname = "max_size",
3414 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3415 .maxlen = sizeof(int),
3417 .proc_handler = proc_dointvec,
3420 .procname = "gc_min_interval",
3421 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3422 .maxlen = sizeof(int),
3424 .proc_handler = proc_dointvec_jiffies,
3427 .procname = "gc_timeout",
3428 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3429 .maxlen = sizeof(int),
3431 .proc_handler = proc_dointvec_jiffies,
3434 .procname = "gc_interval",
3435 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3436 .maxlen = sizeof(int),
3438 .proc_handler = proc_dointvec_jiffies,
3441 .procname = "gc_elasticity",
3442 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3443 .maxlen = sizeof(int),
3445 .proc_handler = proc_dointvec,
3448 .procname = "mtu_expires",
3449 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3450 .maxlen = sizeof(int),
3452 .proc_handler = proc_dointvec_jiffies,
3455 .procname = "min_adv_mss",
3456 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3457 .maxlen = sizeof(int),
3459 .proc_handler = proc_dointvec,
3462 .procname = "gc_min_interval_ms",
3463 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3464 .maxlen = sizeof(int),
3466 .proc_handler = proc_dointvec_ms_jiffies,
3471 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3473 struct ctl_table *table;
3475 table = kmemdup(ipv6_route_table_template,
3476 sizeof(ipv6_route_table_template),
3480 table[0].data = &net->ipv6.sysctl.flush_delay;
3481 table[0].extra1 = net;
3482 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3483 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3484 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3485 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3486 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3487 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3488 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3489 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3490 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3492 /* Don't export sysctls to unprivileged users */
3493 if (net->user_ns != &init_user_ns)
3494 table[0].procname = NULL;
3501 static int __net_init ip6_route_net_init(struct net *net)
3505 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3506 sizeof(net->ipv6.ip6_dst_ops));
3508 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3509 goto out_ip6_dst_ops;
3511 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3512 sizeof(*net->ipv6.ip6_null_entry),
3514 if (!net->ipv6.ip6_null_entry)
3515 goto out_ip6_dst_entries;
3516 net->ipv6.ip6_null_entry->dst.path =
3517 (struct dst_entry *)net->ipv6.ip6_null_entry;
3518 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3519 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3520 ip6_template_metrics, true);
3522 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3523 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3524 sizeof(*net->ipv6.ip6_prohibit_entry),
3526 if (!net->ipv6.ip6_prohibit_entry)
3527 goto out_ip6_null_entry;
3528 net->ipv6.ip6_prohibit_entry->dst.path =
3529 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3530 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3531 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3532 ip6_template_metrics, true);
3534 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3535 sizeof(*net->ipv6.ip6_blk_hole_entry),
3537 if (!net->ipv6.ip6_blk_hole_entry)
3538 goto out_ip6_prohibit_entry;
3539 net->ipv6.ip6_blk_hole_entry->dst.path =
3540 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3541 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3542 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3543 ip6_template_metrics, true);
3546 net->ipv6.sysctl.flush_delay = 0;
3547 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3548 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3549 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3550 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3551 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3552 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3553 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3555 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3561 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3562 out_ip6_prohibit_entry:
3563 kfree(net->ipv6.ip6_prohibit_entry);
3565 kfree(net->ipv6.ip6_null_entry);
3567 out_ip6_dst_entries:
3568 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3573 static void __net_exit ip6_route_net_exit(struct net *net)
3575 kfree(net->ipv6.ip6_null_entry);
3576 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3577 kfree(net->ipv6.ip6_prohibit_entry);
3578 kfree(net->ipv6.ip6_blk_hole_entry);
3580 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3583 static int __net_init ip6_route_net_init_late(struct net *net)
3585 #ifdef CONFIG_PROC_FS
3586 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3587 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3592 static void __net_exit ip6_route_net_exit_late(struct net *net)
3594 #ifdef CONFIG_PROC_FS
3595 remove_proc_entry("ipv6_route", net->proc_net);
3596 remove_proc_entry("rt6_stats", net->proc_net);
3600 static struct pernet_operations ip6_route_net_ops = {
3601 .init = ip6_route_net_init,
3602 .exit = ip6_route_net_exit,
3605 static int __net_init ipv6_inetpeer_init(struct net *net)
3607 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3611 inet_peer_base_init(bp);
3612 net->ipv6.peers = bp;
3616 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3618 struct inet_peer_base *bp = net->ipv6.peers;
3620 net->ipv6.peers = NULL;
3621 inetpeer_invalidate_tree(bp);
3625 static struct pernet_operations ipv6_inetpeer_ops = {
3626 .init = ipv6_inetpeer_init,
3627 .exit = ipv6_inetpeer_exit,
3630 static struct pernet_operations ip6_route_net_late_ops = {
3631 .init = ip6_route_net_init_late,
3632 .exit = ip6_route_net_exit_late,
3635 static struct notifier_block ip6_route_dev_notifier = {
3636 .notifier_call = ip6_route_dev_notify,
3640 int __init ip6_route_init(void)
3646 ip6_dst_ops_template.kmem_cachep =
3647 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3648 SLAB_HWCACHE_ALIGN, NULL);
3649 if (!ip6_dst_ops_template.kmem_cachep)
3652 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3654 goto out_kmem_cache;
3656 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3658 goto out_dst_entries;
3660 ret = register_pernet_subsys(&ip6_route_net_ops);
3662 goto out_register_inetpeer;
3664 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3666 /* Registering of the loopback is done before this portion of code,
3667 * the loopback reference in rt6_info will not be taken, do it
3668 * manually for init_net */
3669 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3670 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3671 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3672 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3673 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3674 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3675 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3679 goto out_register_subsys;
3685 ret = fib6_rules_init();
3689 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3691 goto fib6_rules_init;
3694 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3695 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3696 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3697 goto out_register_late_subsys;
3699 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3701 goto out_register_late_subsys;
3703 for_each_possible_cpu(cpu) {
3704 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3706 INIT_LIST_HEAD(&ul->head);
3707 spin_lock_init(&ul->lock);
3713 out_register_late_subsys:
3714 unregister_pernet_subsys(&ip6_route_net_late_ops);
3716 fib6_rules_cleanup();
3721 out_register_subsys:
3722 unregister_pernet_subsys(&ip6_route_net_ops);
3723 out_register_inetpeer:
3724 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3726 dst_entries_destroy(&ip6_dst_blackhole_ops);
3728 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3732 void ip6_route_cleanup(void)
3734 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3735 unregister_pernet_subsys(&ip6_route_net_late_ops);
3736 fib6_rules_cleanup();
3739 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3740 unregister_pernet_subsys(&ip6_route_net_ops);
3741 dst_entries_destroy(&ip6_dst_blackhole_ops);
3742 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);