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 net *net, 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 net *net, 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_out,
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)
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) ||
1174 flags |= RT6_LOOKUP_F_IFACE;
1176 if (!ipv6_addr_any(&fl6->saddr))
1177 flags |= RT6_LOOKUP_F_HAS_SADDR;
1179 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1181 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1183 EXPORT_SYMBOL(ip6_route_output);
1185 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1187 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1188 struct dst_entry *new = NULL;
1190 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1194 memset(new + 1, 0, sizeof(*rt) - sizeof(*new));
1197 new->input = dst_discard;
1198 new->output = dst_discard_out;
1200 if (dst_metrics_read_only(&ort->dst))
1201 new->_metrics = ort->dst._metrics;
1203 dst_copy_metrics(new, &ort->dst);
1204 rt->rt6i_idev = ort->rt6i_idev;
1206 in6_dev_hold(rt->rt6i_idev);
1208 rt->rt6i_gateway = ort->rt6i_gateway;
1209 rt->rt6i_flags = ort->rt6i_flags;
1210 rt->rt6i_metric = 0;
1212 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1213 #ifdef CONFIG_IPV6_SUBTREES
1214 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1220 dst_release(dst_orig);
1221 return new ? new : ERR_PTR(-ENOMEM);
1225 * Destination cache support functions
1228 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1231 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1232 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1235 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1237 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1240 if (rt6_check_expired(rt))
1246 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1248 if (rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1249 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1255 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1257 struct rt6_info *rt;
1259 rt = (struct rt6_info *) dst;
1261 /* All IPV6 dsts are created with ->obsolete set to the value
1262 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1263 * into this function always.
1266 rt6_dst_from_metrics_check(rt);
1268 if ((rt->rt6i_flags & RTF_PCPU) || unlikely(dst->flags & DST_NOCACHE))
1269 return rt6_dst_from_check(rt, cookie);
1271 return rt6_check(rt, cookie);
1274 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1276 struct rt6_info *rt = (struct rt6_info *) dst;
1279 if (rt->rt6i_flags & RTF_CACHE) {
1280 if (rt6_check_expired(rt)) {
1292 static void ip6_link_failure(struct sk_buff *skb)
1294 struct rt6_info *rt;
1296 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1298 rt = (struct rt6_info *) skb_dst(skb);
1300 if (rt->rt6i_flags & RTF_CACHE) {
1303 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1304 rt->rt6i_node->fn_sernum = -1;
1309 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1311 struct net *net = dev_net(rt->dst.dev);
1313 rt->rt6i_flags |= RTF_MODIFIED;
1314 rt->rt6i_pmtu = mtu;
1315 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1318 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1319 const struct ipv6hdr *iph, u32 mtu)
1321 struct rt6_info *rt6 = (struct rt6_info *)dst;
1323 if (rt6->rt6i_flags & RTF_LOCAL)
1327 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1328 if (mtu >= dst_mtu(dst))
1331 if (rt6->rt6i_flags & RTF_CACHE) {
1332 rt6_do_update_pmtu(rt6, mtu);
1334 const struct in6_addr *daddr, *saddr;
1335 struct rt6_info *nrt6;
1338 daddr = &iph->daddr;
1339 saddr = &iph->saddr;
1341 daddr = &sk->sk_v6_daddr;
1342 saddr = &inet6_sk(sk)->saddr;
1346 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1348 rt6_do_update_pmtu(nrt6, mtu);
1350 /* ip6_ins_rt(nrt6) will bump the
1351 * rt6->rt6i_node->fn_sernum
1352 * which will fail the next rt6_check() and
1353 * invalidate the sk->sk_dst_cache.
1360 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1361 struct sk_buff *skb, u32 mtu)
1363 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1366 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1369 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1370 struct dst_entry *dst;
1373 memset(&fl6, 0, sizeof(fl6));
1374 fl6.flowi6_oif = oif;
1375 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1376 fl6.daddr = iph->daddr;
1377 fl6.saddr = iph->saddr;
1378 fl6.flowlabel = ip6_flowinfo(iph);
1380 dst = ip6_route_output(net, NULL, &fl6);
1382 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1385 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1387 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1389 ip6_update_pmtu(skb, sock_net(sk), mtu,
1390 sk->sk_bound_dev_if, sk->sk_mark);
1392 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1394 /* Handle redirects */
1395 struct ip6rd_flowi {
1397 struct in6_addr gateway;
1400 static struct rt6_info *__ip6_route_redirect(struct net *net,
1401 struct fib6_table *table,
1405 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1406 struct rt6_info *rt;
1407 struct fib6_node *fn;
1409 /* Get the "current" route for this destination and
1410 * check if the redirect has come from approriate router.
1412 * RFC 4861 specifies that redirects should only be
1413 * accepted if they come from the nexthop to the target.
1414 * Due to the way the routes are chosen, this notion
1415 * is a bit fuzzy and one might need to check all possible
1419 read_lock_bh(&table->tb6_lock);
1420 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1422 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1423 if (rt6_check_expired(rt))
1427 if (!(rt->rt6i_flags & RTF_GATEWAY))
1429 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1431 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1437 rt = net->ipv6.ip6_null_entry;
1438 else if (rt->dst.error) {
1439 rt = net->ipv6.ip6_null_entry;
1443 if (rt == net->ipv6.ip6_null_entry) {
1444 fn = fib6_backtrack(fn, &fl6->saddr);
1452 read_unlock_bh(&table->tb6_lock);
1457 static struct dst_entry *ip6_route_redirect(struct net *net,
1458 const struct flowi6 *fl6,
1459 const struct in6_addr *gateway)
1461 int flags = RT6_LOOKUP_F_HAS_SADDR;
1462 struct ip6rd_flowi rdfl;
1465 rdfl.gateway = *gateway;
1467 return fib6_rule_lookup(net, &rdfl.fl6,
1468 flags, __ip6_route_redirect);
1471 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1473 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1474 struct dst_entry *dst;
1477 memset(&fl6, 0, sizeof(fl6));
1478 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1479 fl6.flowi6_oif = oif;
1480 fl6.flowi6_mark = mark;
1481 fl6.daddr = iph->daddr;
1482 fl6.saddr = iph->saddr;
1483 fl6.flowlabel = ip6_flowinfo(iph);
1485 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1486 rt6_do_redirect(dst, NULL, skb);
1489 EXPORT_SYMBOL_GPL(ip6_redirect);
1491 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1494 const struct ipv6hdr *iph = ipv6_hdr(skb);
1495 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1496 struct dst_entry *dst;
1499 memset(&fl6, 0, sizeof(fl6));
1500 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1501 fl6.flowi6_oif = oif;
1502 fl6.flowi6_mark = mark;
1503 fl6.daddr = msg->dest;
1504 fl6.saddr = iph->daddr;
1506 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1507 rt6_do_redirect(dst, NULL, skb);
1511 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1513 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1515 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1517 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1519 struct net_device *dev = dst->dev;
1520 unsigned int mtu = dst_mtu(dst);
1521 struct net *net = dev_net(dev);
1523 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1525 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1526 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1529 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1530 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1531 * IPV6_MAXPLEN is also valid and means: "any MSS,
1532 * rely only on pmtu discovery"
1534 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1539 static unsigned int ip6_mtu(const struct dst_entry *dst)
1541 const struct rt6_info *rt = (const struct rt6_info *)dst;
1542 unsigned int mtu = rt->rt6i_pmtu;
1543 struct inet6_dev *idev;
1548 mtu = dst_metric_raw(dst, RTAX_MTU);
1555 idev = __in6_dev_get(dst->dev);
1557 mtu = idev->cnf.mtu6;
1561 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1564 static struct dst_entry *icmp6_dst_gc_list;
1565 static DEFINE_SPINLOCK(icmp6_dst_lock);
1567 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1570 struct dst_entry *dst;
1571 struct rt6_info *rt;
1572 struct inet6_dev *idev = in6_dev_get(dev);
1573 struct net *net = dev_net(dev);
1575 if (unlikely(!idev))
1576 return ERR_PTR(-ENODEV);
1578 rt = ip6_dst_alloc(net, dev, 0);
1579 if (unlikely(!rt)) {
1581 dst = ERR_PTR(-ENOMEM);
1585 rt->dst.flags |= DST_HOST;
1586 rt->dst.output = ip6_output;
1587 atomic_set(&rt->dst.__refcnt, 1);
1588 rt->rt6i_gateway = fl6->daddr;
1589 rt->rt6i_dst.addr = fl6->daddr;
1590 rt->rt6i_dst.plen = 128;
1591 rt->rt6i_idev = idev;
1592 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1594 spin_lock_bh(&icmp6_dst_lock);
1595 rt->dst.next = icmp6_dst_gc_list;
1596 icmp6_dst_gc_list = &rt->dst;
1597 spin_unlock_bh(&icmp6_dst_lock);
1599 fib6_force_start_gc(net);
1601 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1607 int icmp6_dst_gc(void)
1609 struct dst_entry *dst, **pprev;
1612 spin_lock_bh(&icmp6_dst_lock);
1613 pprev = &icmp6_dst_gc_list;
1615 while ((dst = *pprev) != NULL) {
1616 if (!atomic_read(&dst->__refcnt)) {
1625 spin_unlock_bh(&icmp6_dst_lock);
1630 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1633 struct dst_entry *dst, **pprev;
1635 spin_lock_bh(&icmp6_dst_lock);
1636 pprev = &icmp6_dst_gc_list;
1637 while ((dst = *pprev) != NULL) {
1638 struct rt6_info *rt = (struct rt6_info *) dst;
1639 if (func(rt, arg)) {
1646 spin_unlock_bh(&icmp6_dst_lock);
1649 static int ip6_dst_gc(struct dst_ops *ops)
1651 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1652 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1653 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1654 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1655 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1656 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1659 entries = dst_entries_get_fast(ops);
1660 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1661 entries <= rt_max_size)
1664 net->ipv6.ip6_rt_gc_expire++;
1665 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1666 entries = dst_entries_get_slow(ops);
1667 if (entries < ops->gc_thresh)
1668 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1670 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1671 return entries > rt_max_size;
1674 static int ip6_convert_metrics(struct mx6_config *mxc,
1675 const struct fib6_config *cfg)
1677 bool ecn_ca = false;
1685 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1689 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1690 int type = nla_type(nla);
1695 if (unlikely(type > RTAX_MAX))
1698 if (type == RTAX_CC_ALGO) {
1699 char tmp[TCP_CA_NAME_MAX];
1701 nla_strlcpy(tmp, nla, sizeof(tmp));
1702 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1703 if (val == TCP_CA_UNSPEC)
1706 val = nla_get_u32(nla);
1708 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1712 __set_bit(type - 1, mxc->mx_valid);
1716 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1717 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1727 int ip6_route_info_create(struct fib6_config *cfg, struct rt6_info **rt_ret)
1730 struct net *net = cfg->fc_nlinfo.nl_net;
1731 struct rt6_info *rt = NULL;
1732 struct net_device *dev = NULL;
1733 struct inet6_dev *idev = NULL;
1734 struct fib6_table *table;
1737 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1739 #ifndef CONFIG_IPV6_SUBTREES
1740 if (cfg->fc_src_len)
1743 if (cfg->fc_ifindex) {
1745 dev = dev_get_by_index(net, cfg->fc_ifindex);
1748 idev = in6_dev_get(dev);
1753 if (cfg->fc_metric == 0)
1754 cfg->fc_metric = IP6_RT_PRIO_USER;
1757 if (cfg->fc_nlinfo.nlh &&
1758 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1759 table = fib6_get_table(net, cfg->fc_table);
1761 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1762 table = fib6_new_table(net, cfg->fc_table);
1765 table = fib6_new_table(net, cfg->fc_table);
1771 rt = ip6_dst_alloc(net, NULL,
1772 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1779 if (cfg->fc_flags & RTF_EXPIRES)
1780 rt6_set_expires(rt, jiffies +
1781 clock_t_to_jiffies(cfg->fc_expires));
1783 rt6_clean_expires(rt);
1785 if (cfg->fc_protocol == RTPROT_UNSPEC)
1786 cfg->fc_protocol = RTPROT_BOOT;
1787 rt->rt6i_protocol = cfg->fc_protocol;
1789 addr_type = ipv6_addr_type(&cfg->fc_dst);
1791 if (addr_type & IPV6_ADDR_MULTICAST)
1792 rt->dst.input = ip6_mc_input;
1793 else if (cfg->fc_flags & RTF_LOCAL)
1794 rt->dst.input = ip6_input;
1796 rt->dst.input = ip6_forward;
1798 rt->dst.output = ip6_output;
1800 if (cfg->fc_encap) {
1801 struct lwtunnel_state *lwtstate;
1803 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1804 cfg->fc_encap, AF_INET6, cfg,
1808 rt->dst.lwtstate = lwtstate_get(lwtstate);
1809 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1810 rt->dst.lwtstate->orig_output = rt->dst.output;
1811 rt->dst.output = lwtunnel_output;
1813 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1814 rt->dst.lwtstate->orig_input = rt->dst.input;
1815 rt->dst.input = lwtunnel_input;
1819 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1820 rt->rt6i_dst.plen = cfg->fc_dst_len;
1821 if (rt->rt6i_dst.plen == 128)
1822 rt->dst.flags |= DST_HOST;
1824 #ifdef CONFIG_IPV6_SUBTREES
1825 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1826 rt->rt6i_src.plen = cfg->fc_src_len;
1829 rt->rt6i_metric = cfg->fc_metric;
1831 /* We cannot add true routes via loopback here,
1832 they would result in kernel looping; promote them to reject routes
1834 if ((cfg->fc_flags & RTF_REJECT) ||
1835 (dev && (dev->flags & IFF_LOOPBACK) &&
1836 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1837 !(cfg->fc_flags & RTF_LOCAL))) {
1838 /* hold loopback dev/idev if we haven't done so. */
1839 if (dev != net->loopback_dev) {
1844 dev = net->loopback_dev;
1846 idev = in6_dev_get(dev);
1852 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1853 switch (cfg->fc_type) {
1855 rt->dst.error = -EINVAL;
1856 rt->dst.output = dst_discard_out;
1857 rt->dst.input = dst_discard;
1860 rt->dst.error = -EACCES;
1861 rt->dst.output = ip6_pkt_prohibit_out;
1862 rt->dst.input = ip6_pkt_prohibit;
1865 case RTN_UNREACHABLE:
1867 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1868 : (cfg->fc_type == RTN_UNREACHABLE)
1869 ? -EHOSTUNREACH : -ENETUNREACH;
1870 rt->dst.output = ip6_pkt_discard_out;
1871 rt->dst.input = ip6_pkt_discard;
1877 if (cfg->fc_flags & RTF_GATEWAY) {
1878 const struct in6_addr *gw_addr;
1881 gw_addr = &cfg->fc_gateway;
1882 gwa_type = ipv6_addr_type(gw_addr);
1884 /* if gw_addr is local we will fail to detect this in case
1885 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1886 * will return already-added prefix route via interface that
1887 * prefix route was assigned to, which might be non-loopback.
1890 if (ipv6_chk_addr_and_flags(net, gw_addr,
1891 gwa_type & IPV6_ADDR_LINKLOCAL ?
1895 rt->rt6i_gateway = *gw_addr;
1897 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1898 struct rt6_info *grt;
1900 /* IPv6 strictly inhibits using not link-local
1901 addresses as nexthop address.
1902 Otherwise, router will not able to send redirects.
1903 It is very good, but in some (rare!) circumstances
1904 (SIT, PtP, NBMA NOARP links) it is handy to allow
1905 some exceptions. --ANK
1907 if (!(gwa_type & IPV6_ADDR_UNICAST))
1910 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1912 err = -EHOSTUNREACH;
1916 if (dev != grt->dst.dev) {
1922 idev = grt->rt6i_idev;
1924 in6_dev_hold(grt->rt6i_idev);
1926 if (!(grt->rt6i_flags & RTF_GATEWAY))
1934 if (!dev || (dev->flags & IFF_LOOPBACK))
1942 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1943 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1947 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1948 rt->rt6i_prefsrc.plen = 128;
1950 rt->rt6i_prefsrc.plen = 0;
1952 rt->rt6i_flags = cfg->fc_flags;
1956 rt->rt6i_idev = idev;
1957 rt->rt6i_table = table;
1959 cfg->fc_nlinfo.nl_net = dev_net(dev);
1977 int ip6_route_add(struct fib6_config *cfg)
1979 struct mx6_config mxc = { .mx = NULL, };
1980 struct rt6_info *rt = NULL;
1983 err = ip6_route_info_create(cfg, &rt);
1987 err = ip6_convert_metrics(&mxc, cfg);
1991 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2003 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2006 struct fib6_table *table;
2007 struct net *net = dev_net(rt->dst.dev);
2009 if (rt == net->ipv6.ip6_null_entry ||
2010 rt->dst.flags & DST_NOCACHE) {
2015 table = rt->rt6i_table;
2016 write_lock_bh(&table->tb6_lock);
2017 err = fib6_del(rt, info);
2018 write_unlock_bh(&table->tb6_lock);
2025 int ip6_del_rt(struct rt6_info *rt)
2027 struct nl_info info = {
2028 .nl_net = dev_net(rt->dst.dev),
2030 return __ip6_del_rt(rt, &info);
2033 static int ip6_route_del(struct fib6_config *cfg)
2035 struct fib6_table *table;
2036 struct fib6_node *fn;
2037 struct rt6_info *rt;
2040 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2044 read_lock_bh(&table->tb6_lock);
2046 fn = fib6_locate(&table->tb6_root,
2047 &cfg->fc_dst, cfg->fc_dst_len,
2048 &cfg->fc_src, cfg->fc_src_len);
2051 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2052 if ((rt->rt6i_flags & RTF_CACHE) &&
2053 !(cfg->fc_flags & RTF_CACHE))
2055 if (cfg->fc_ifindex &&
2057 rt->dst.dev->ifindex != cfg->fc_ifindex))
2059 if (cfg->fc_flags & RTF_GATEWAY &&
2060 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2062 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2065 read_unlock_bh(&table->tb6_lock);
2067 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2070 read_unlock_bh(&table->tb6_lock);
2075 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2077 struct net *net = dev_net(skb->dev);
2078 struct netevent_redirect netevent;
2079 struct rt6_info *rt, *nrt = NULL;
2080 struct ndisc_options ndopts;
2081 struct inet6_dev *in6_dev;
2082 struct neighbour *neigh;
2084 int optlen, on_link;
2087 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2088 optlen -= sizeof(*msg);
2091 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2095 msg = (struct rd_msg *)icmp6_hdr(skb);
2097 if (ipv6_addr_is_multicast(&msg->dest)) {
2098 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2103 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2105 } else if (ipv6_addr_type(&msg->target) !=
2106 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2107 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2111 in6_dev = __in6_dev_get(skb->dev);
2114 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2118 * The IP source address of the Redirect MUST be the same as the current
2119 * first-hop router for the specified ICMP Destination Address.
2122 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2123 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2128 if (ndopts.nd_opts_tgt_lladdr) {
2129 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2132 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2137 rt = (struct rt6_info *) dst;
2138 if (rt == net->ipv6.ip6_null_entry) {
2139 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2143 /* Redirect received -> path was valid.
2144 * Look, redirects are sent only in response to data packets,
2145 * so that this nexthop apparently is reachable. --ANK
2147 dst_confirm(&rt->dst);
2149 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2154 * We have finally decided to accept it.
2157 neigh_update(neigh, lladdr, NUD_STALE,
2158 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2159 NEIGH_UPDATE_F_OVERRIDE|
2160 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2161 NEIGH_UPDATE_F_ISROUTER))
2164 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2168 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2170 nrt->rt6i_flags &= ~RTF_GATEWAY;
2172 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2174 if (ip6_ins_rt(nrt))
2177 netevent.old = &rt->dst;
2178 netevent.new = &nrt->dst;
2179 netevent.daddr = &msg->dest;
2180 netevent.neigh = neigh;
2181 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2183 if (rt->rt6i_flags & RTF_CACHE) {
2184 rt = (struct rt6_info *) dst_clone(&rt->dst);
2189 neigh_release(neigh);
2193 * Misc support functions
2196 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2198 BUG_ON(from->dst.from);
2200 rt->rt6i_flags &= ~RTF_EXPIRES;
2201 dst_hold(&from->dst);
2202 rt->dst.from = &from->dst;
2203 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2206 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2208 rt->dst.input = ort->dst.input;
2209 rt->dst.output = ort->dst.output;
2210 rt->rt6i_dst = ort->rt6i_dst;
2211 rt->dst.error = ort->dst.error;
2212 rt->rt6i_idev = ort->rt6i_idev;
2214 in6_dev_hold(rt->rt6i_idev);
2215 rt->dst.lastuse = jiffies;
2216 rt->rt6i_gateway = ort->rt6i_gateway;
2217 rt->rt6i_flags = ort->rt6i_flags;
2218 rt6_set_from(rt, ort);
2219 rt->rt6i_metric = ort->rt6i_metric;
2220 #ifdef CONFIG_IPV6_SUBTREES
2221 rt->rt6i_src = ort->rt6i_src;
2223 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2224 rt->rt6i_table = ort->rt6i_table;
2225 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2228 #ifdef CONFIG_IPV6_ROUTE_INFO
2229 static struct rt6_info *rt6_get_route_info(struct net *net,
2230 const struct in6_addr *prefix, int prefixlen,
2231 const struct in6_addr *gwaddr, int ifindex)
2233 struct fib6_node *fn;
2234 struct rt6_info *rt = NULL;
2235 struct fib6_table *table;
2237 table = fib6_get_table(net, RT6_TABLE_INFO);
2241 read_lock_bh(&table->tb6_lock);
2242 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2246 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2247 if (rt->dst.dev->ifindex != ifindex)
2249 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2251 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2257 read_unlock_bh(&table->tb6_lock);
2261 static struct rt6_info *rt6_add_route_info(struct net *net,
2262 const struct in6_addr *prefix, int prefixlen,
2263 const struct in6_addr *gwaddr, int ifindex,
2266 struct fib6_config cfg = {
2267 .fc_table = RT6_TABLE_INFO,
2268 .fc_metric = IP6_RT_PRIO_USER,
2269 .fc_ifindex = ifindex,
2270 .fc_dst_len = prefixlen,
2271 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2272 RTF_UP | RTF_PREF(pref),
2273 .fc_nlinfo.portid = 0,
2274 .fc_nlinfo.nlh = NULL,
2275 .fc_nlinfo.nl_net = net,
2278 cfg.fc_dst = *prefix;
2279 cfg.fc_gateway = *gwaddr;
2281 /* We should treat it as a default route if prefix length is 0. */
2283 cfg.fc_flags |= RTF_DEFAULT;
2285 ip6_route_add(&cfg);
2287 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2291 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2293 struct rt6_info *rt;
2294 struct fib6_table *table;
2296 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2300 read_lock_bh(&table->tb6_lock);
2301 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2302 if (dev == rt->dst.dev &&
2303 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2304 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2309 read_unlock_bh(&table->tb6_lock);
2313 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2314 struct net_device *dev,
2317 struct fib6_config cfg = {
2318 .fc_table = RT6_TABLE_DFLT,
2319 .fc_metric = IP6_RT_PRIO_USER,
2320 .fc_ifindex = dev->ifindex,
2321 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2322 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2323 .fc_nlinfo.portid = 0,
2324 .fc_nlinfo.nlh = NULL,
2325 .fc_nlinfo.nl_net = dev_net(dev),
2328 cfg.fc_gateway = *gwaddr;
2330 ip6_route_add(&cfg);
2332 return rt6_get_dflt_router(gwaddr, dev);
2335 void rt6_purge_dflt_routers(struct net *net)
2337 struct rt6_info *rt;
2338 struct fib6_table *table;
2340 /* NOTE: Keep consistent with rt6_get_dflt_router */
2341 table = fib6_get_table(net, RT6_TABLE_DFLT);
2346 read_lock_bh(&table->tb6_lock);
2347 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2348 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2349 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2351 read_unlock_bh(&table->tb6_lock);
2356 read_unlock_bh(&table->tb6_lock);
2359 static void rtmsg_to_fib6_config(struct net *net,
2360 struct in6_rtmsg *rtmsg,
2361 struct fib6_config *cfg)
2363 memset(cfg, 0, sizeof(*cfg));
2365 cfg->fc_table = RT6_TABLE_MAIN;
2366 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2367 cfg->fc_metric = rtmsg->rtmsg_metric;
2368 cfg->fc_expires = rtmsg->rtmsg_info;
2369 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2370 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2371 cfg->fc_flags = rtmsg->rtmsg_flags;
2373 cfg->fc_nlinfo.nl_net = net;
2375 cfg->fc_dst = rtmsg->rtmsg_dst;
2376 cfg->fc_src = rtmsg->rtmsg_src;
2377 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2380 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2382 struct fib6_config cfg;
2383 struct in6_rtmsg rtmsg;
2387 case SIOCADDRT: /* Add a route */
2388 case SIOCDELRT: /* Delete a route */
2389 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2391 err = copy_from_user(&rtmsg, arg,
2392 sizeof(struct in6_rtmsg));
2396 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2401 err = ip6_route_add(&cfg);
2404 err = ip6_route_del(&cfg);
2418 * Drop the packet on the floor
2421 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2424 struct dst_entry *dst = skb_dst(skb);
2425 switch (ipstats_mib_noroutes) {
2426 case IPSTATS_MIB_INNOROUTES:
2427 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2428 if (type == IPV6_ADDR_ANY) {
2429 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2430 IPSTATS_MIB_INADDRERRORS);
2434 case IPSTATS_MIB_OUTNOROUTES:
2435 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2436 ipstats_mib_noroutes);
2439 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2444 static int ip6_pkt_discard(struct sk_buff *skb)
2446 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2449 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2451 skb->dev = skb_dst(skb)->dev;
2452 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2455 static int ip6_pkt_prohibit(struct sk_buff *skb)
2457 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2460 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2462 skb->dev = skb_dst(skb)->dev;
2463 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2467 * Allocate a dst for local (unicast / anycast) address.
2470 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2471 const struct in6_addr *addr,
2474 struct net *net = dev_net(idev->dev);
2475 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2478 return ERR_PTR(-ENOMEM);
2482 rt->dst.flags |= DST_HOST;
2483 rt->dst.input = ip6_input;
2484 rt->dst.output = ip6_output;
2485 rt->rt6i_idev = idev;
2487 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2489 rt->rt6i_flags |= RTF_ANYCAST;
2491 rt->rt6i_flags |= RTF_LOCAL;
2493 rt->rt6i_gateway = *addr;
2494 rt->rt6i_dst.addr = *addr;
2495 rt->rt6i_dst.plen = 128;
2496 rt->rt6i_table = fib6_get_table(net, RT6_TABLE_LOCAL);
2497 rt->dst.flags |= DST_NOCACHE;
2499 atomic_set(&rt->dst.__refcnt, 1);
2504 int ip6_route_get_saddr(struct net *net,
2505 struct rt6_info *rt,
2506 const struct in6_addr *daddr,
2508 struct in6_addr *saddr)
2510 struct inet6_dev *idev =
2511 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2513 if (rt && rt->rt6i_prefsrc.plen)
2514 *saddr = rt->rt6i_prefsrc.addr;
2516 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2517 daddr, prefs, saddr);
2521 /* remove deleted ip from prefsrc entries */
2522 struct arg_dev_net_ip {
2523 struct net_device *dev;
2525 struct in6_addr *addr;
2528 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2530 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2531 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2532 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2534 if (((void *)rt->dst.dev == dev || !dev) &&
2535 rt != net->ipv6.ip6_null_entry &&
2536 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2537 /* remove prefsrc entry */
2538 rt->rt6i_prefsrc.plen = 0;
2543 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2545 struct net *net = dev_net(ifp->idev->dev);
2546 struct arg_dev_net_ip adni = {
2547 .dev = ifp->idev->dev,
2551 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2554 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2555 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2557 /* Remove routers and update dst entries when gateway turn into host. */
2558 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2560 struct in6_addr *gateway = (struct in6_addr *)arg;
2562 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2563 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2564 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2570 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2572 fib6_clean_all(net, fib6_clean_tohost, gateway);
2575 struct arg_dev_net {
2576 struct net_device *dev;
2580 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2582 const struct arg_dev_net *adn = arg;
2583 const struct net_device *dev = adn->dev;
2585 if ((rt->dst.dev == dev || !dev) &&
2586 rt != adn->net->ipv6.ip6_null_entry)
2592 void rt6_ifdown(struct net *net, struct net_device *dev)
2594 struct arg_dev_net adn = {
2599 fib6_clean_all(net, fib6_ifdown, &adn);
2600 icmp6_clean_all(fib6_ifdown, &adn);
2601 rt6_uncached_list_flush_dev(net, dev);
2604 struct rt6_mtu_change_arg {
2605 struct net_device *dev;
2609 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2611 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2612 struct inet6_dev *idev;
2614 /* In IPv6 pmtu discovery is not optional,
2615 so that RTAX_MTU lock cannot disable it.
2616 We still use this lock to block changes
2617 caused by addrconf/ndisc.
2620 idev = __in6_dev_get(arg->dev);
2624 /* For administrative MTU increase, there is no way to discover
2625 IPv6 PMTU increase, so PMTU increase should be updated here.
2626 Since RFC 1981 doesn't include administrative MTU increase
2627 update PMTU increase is a MUST. (i.e. jumbo frame)
2630 If new MTU is less than route PMTU, this new MTU will be the
2631 lowest MTU in the path, update the route PMTU to reflect PMTU
2632 decreases; if new MTU is greater than route PMTU, and the
2633 old MTU is the lowest MTU in the path, update the route PMTU
2634 to reflect the increase. In this case if the other nodes' MTU
2635 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2638 if (rt->dst.dev == arg->dev &&
2639 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2640 if (rt->rt6i_flags & RTF_CACHE) {
2641 /* For RTF_CACHE with rt6i_pmtu == 0
2642 * (i.e. a redirected route),
2643 * the metrics of its rt->dst.from has already
2646 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2647 rt->rt6i_pmtu = arg->mtu;
2648 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2649 (dst_mtu(&rt->dst) < arg->mtu &&
2650 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2651 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2657 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2659 struct rt6_mtu_change_arg arg = {
2664 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2667 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2668 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2669 [RTA_OIF] = { .type = NLA_U32 },
2670 [RTA_IIF] = { .type = NLA_U32 },
2671 [RTA_PRIORITY] = { .type = NLA_U32 },
2672 [RTA_METRICS] = { .type = NLA_NESTED },
2673 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2674 [RTA_PREF] = { .type = NLA_U8 },
2675 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2676 [RTA_ENCAP] = { .type = NLA_NESTED },
2679 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2680 struct fib6_config *cfg)
2683 struct nlattr *tb[RTA_MAX+1];
2687 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2692 rtm = nlmsg_data(nlh);
2693 memset(cfg, 0, sizeof(*cfg));
2695 cfg->fc_table = rtm->rtm_table;
2696 cfg->fc_dst_len = rtm->rtm_dst_len;
2697 cfg->fc_src_len = rtm->rtm_src_len;
2698 cfg->fc_flags = RTF_UP;
2699 cfg->fc_protocol = rtm->rtm_protocol;
2700 cfg->fc_type = rtm->rtm_type;
2702 if (rtm->rtm_type == RTN_UNREACHABLE ||
2703 rtm->rtm_type == RTN_BLACKHOLE ||
2704 rtm->rtm_type == RTN_PROHIBIT ||
2705 rtm->rtm_type == RTN_THROW)
2706 cfg->fc_flags |= RTF_REJECT;
2708 if (rtm->rtm_type == RTN_LOCAL)
2709 cfg->fc_flags |= RTF_LOCAL;
2711 if (rtm->rtm_flags & RTM_F_CLONED)
2712 cfg->fc_flags |= RTF_CACHE;
2714 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2715 cfg->fc_nlinfo.nlh = nlh;
2716 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2718 if (tb[RTA_GATEWAY]) {
2719 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2720 cfg->fc_flags |= RTF_GATEWAY;
2724 int plen = (rtm->rtm_dst_len + 7) >> 3;
2726 if (nla_len(tb[RTA_DST]) < plen)
2729 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2733 int plen = (rtm->rtm_src_len + 7) >> 3;
2735 if (nla_len(tb[RTA_SRC]) < plen)
2738 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2741 if (tb[RTA_PREFSRC])
2742 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2745 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2747 if (tb[RTA_PRIORITY])
2748 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2750 if (tb[RTA_METRICS]) {
2751 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2752 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2756 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2758 if (tb[RTA_MULTIPATH]) {
2759 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2760 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2764 pref = nla_get_u8(tb[RTA_PREF]);
2765 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2766 pref != ICMPV6_ROUTER_PREF_HIGH)
2767 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2768 cfg->fc_flags |= RTF_PREF(pref);
2772 cfg->fc_encap = tb[RTA_ENCAP];
2774 if (tb[RTA_ENCAP_TYPE])
2775 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2783 struct rt6_info *rt6_info;
2784 struct fib6_config r_cfg;
2785 struct mx6_config mxc;
2786 struct list_head next;
2789 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2793 list_for_each_entry(nh, rt6_nh_list, next) {
2794 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2795 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2796 nh->r_cfg.fc_ifindex);
2800 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2801 struct rt6_info *rt, struct fib6_config *r_cfg)
2804 struct rt6_info *rtnh;
2807 list_for_each_entry(nh, rt6_nh_list, next) {
2808 /* check if rt6_info already exists */
2809 rtnh = nh->rt6_info;
2811 if (rtnh->dst.dev == rt->dst.dev &&
2812 rtnh->rt6i_idev == rt->rt6i_idev &&
2813 ipv6_addr_equal(&rtnh->rt6i_gateway,
2818 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2822 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2827 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2828 list_add_tail(&nh->next, rt6_nh_list);
2833 static int ip6_route_multipath_add(struct fib6_config *cfg)
2835 struct fib6_config r_cfg;
2836 struct rtnexthop *rtnh;
2837 struct rt6_info *rt;
2838 struct rt6_nh *err_nh;
2839 struct rt6_nh *nh, *nh_safe;
2844 int replace = (cfg->fc_nlinfo.nlh &&
2845 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2846 LIST_HEAD(rt6_nh_list);
2848 remaining = cfg->fc_mp_len;
2849 rtnh = (struct rtnexthop *)cfg->fc_mp;
2851 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2852 * rt6_info structs per nexthop
2854 while (rtnh_ok(rtnh, remaining)) {
2855 memcpy(&r_cfg, cfg, sizeof(*cfg));
2856 if (rtnh->rtnh_ifindex)
2857 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2859 attrlen = rtnh_attrlen(rtnh);
2861 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2863 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2865 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2866 r_cfg.fc_flags |= RTF_GATEWAY;
2868 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2869 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2871 r_cfg.fc_encap_type = nla_get_u16(nla);
2874 err = ip6_route_info_create(&r_cfg, &rt);
2878 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2884 rtnh = rtnh_next(rtnh, &remaining);
2888 list_for_each_entry(nh, &rt6_nh_list, next) {
2889 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2890 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2891 nh->rt6_info = NULL;
2894 ip6_print_replace_route_err(&rt6_nh_list);
2899 /* Because each route is added like a single route we remove
2900 * these flags after the first nexthop: if there is a collision,
2901 * we have already failed to add the first nexthop:
2902 * fib6_add_rt2node() has rejected it; when replacing, old
2903 * nexthops have been replaced by first new, the rest should
2906 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2914 /* Delete routes that were already added */
2915 list_for_each_entry(nh, &rt6_nh_list, next) {
2918 ip6_route_del(&nh->r_cfg);
2922 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2924 dst_free(&nh->rt6_info->dst);
2926 list_del(&nh->next);
2933 static int ip6_route_multipath_del(struct fib6_config *cfg)
2935 struct fib6_config r_cfg;
2936 struct rtnexthop *rtnh;
2939 int err = 1, last_err = 0;
2941 remaining = cfg->fc_mp_len;
2942 rtnh = (struct rtnexthop *)cfg->fc_mp;
2944 /* Parse a Multipath Entry */
2945 while (rtnh_ok(rtnh, remaining)) {
2946 memcpy(&r_cfg, cfg, sizeof(*cfg));
2947 if (rtnh->rtnh_ifindex)
2948 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2950 attrlen = rtnh_attrlen(rtnh);
2952 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2954 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2956 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2957 r_cfg.fc_flags |= RTF_GATEWAY;
2960 err = ip6_route_del(&r_cfg);
2964 rtnh = rtnh_next(rtnh, &remaining);
2970 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2972 struct fib6_config cfg;
2975 err = rtm_to_fib6_config(skb, nlh, &cfg);
2980 return ip6_route_multipath_del(&cfg);
2982 return ip6_route_del(&cfg);
2985 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2987 struct fib6_config cfg;
2990 err = rtm_to_fib6_config(skb, nlh, &cfg);
2995 return ip6_route_multipath_add(&cfg);
2997 return ip6_route_add(&cfg);
3000 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3002 return NLMSG_ALIGN(sizeof(struct rtmsg))
3003 + nla_total_size(16) /* RTA_SRC */
3004 + nla_total_size(16) /* RTA_DST */
3005 + nla_total_size(16) /* RTA_GATEWAY */
3006 + nla_total_size(16) /* RTA_PREFSRC */
3007 + nla_total_size(4) /* RTA_TABLE */
3008 + nla_total_size(4) /* RTA_IIF */
3009 + nla_total_size(4) /* RTA_OIF */
3010 + nla_total_size(4) /* RTA_PRIORITY */
3011 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3012 + nla_total_size(sizeof(struct rta_cacheinfo))
3013 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3014 + nla_total_size(1) /* RTA_PREF */
3015 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3018 static int rt6_fill_node(struct net *net,
3019 struct sk_buff *skb, struct rt6_info *rt,
3020 struct in6_addr *dst, struct in6_addr *src,
3021 int iif, int type, u32 portid, u32 seq,
3022 int prefix, int nowait, unsigned int flags)
3024 u32 metrics[RTAX_MAX];
3026 struct nlmsghdr *nlh;
3030 if (prefix) { /* user wants prefix routes only */
3031 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3032 /* success since this is not a prefix route */
3037 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3041 rtm = nlmsg_data(nlh);
3042 rtm->rtm_family = AF_INET6;
3043 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3044 rtm->rtm_src_len = rt->rt6i_src.plen;
3047 table = rt->rt6i_table->tb6_id;
3049 table = RT6_TABLE_UNSPEC;
3050 rtm->rtm_table = table;
3051 if (nla_put_u32(skb, RTA_TABLE, table))
3052 goto nla_put_failure;
3053 if (rt->rt6i_flags & RTF_REJECT) {
3054 switch (rt->dst.error) {
3056 rtm->rtm_type = RTN_BLACKHOLE;
3059 rtm->rtm_type = RTN_PROHIBIT;
3062 rtm->rtm_type = RTN_THROW;
3065 rtm->rtm_type = RTN_UNREACHABLE;
3069 else if (rt->rt6i_flags & RTF_LOCAL)
3070 rtm->rtm_type = RTN_LOCAL;
3071 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3072 rtm->rtm_type = RTN_LOCAL;
3074 rtm->rtm_type = RTN_UNICAST;
3076 if (!netif_carrier_ok(rt->dst.dev)) {
3077 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3078 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3079 rtm->rtm_flags |= RTNH_F_DEAD;
3081 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3082 rtm->rtm_protocol = rt->rt6i_protocol;
3083 if (rt->rt6i_flags & RTF_DYNAMIC)
3084 rtm->rtm_protocol = RTPROT_REDIRECT;
3085 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3086 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3087 rtm->rtm_protocol = RTPROT_RA;
3089 rtm->rtm_protocol = RTPROT_KERNEL;
3092 if (rt->rt6i_flags & RTF_CACHE)
3093 rtm->rtm_flags |= RTM_F_CLONED;
3096 if (nla_put_in6_addr(skb, RTA_DST, dst))
3097 goto nla_put_failure;
3098 rtm->rtm_dst_len = 128;
3099 } else if (rtm->rtm_dst_len)
3100 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3101 goto nla_put_failure;
3102 #ifdef CONFIG_IPV6_SUBTREES
3104 if (nla_put_in6_addr(skb, RTA_SRC, src))
3105 goto nla_put_failure;
3106 rtm->rtm_src_len = 128;
3107 } else if (rtm->rtm_src_len &&
3108 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3109 goto nla_put_failure;
3112 #ifdef CONFIG_IPV6_MROUTE
3113 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3114 int err = ip6mr_get_route(net, skb, rtm, nowait);
3119 goto nla_put_failure;
3121 if (err == -EMSGSIZE)
3122 goto nla_put_failure;
3127 if (nla_put_u32(skb, RTA_IIF, iif))
3128 goto nla_put_failure;
3130 struct in6_addr saddr_buf;
3131 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3132 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3133 goto nla_put_failure;
3136 if (rt->rt6i_prefsrc.plen) {
3137 struct in6_addr saddr_buf;
3138 saddr_buf = rt->rt6i_prefsrc.addr;
3139 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3140 goto nla_put_failure;
3143 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3145 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3146 if (rtnetlink_put_metrics(skb, metrics) < 0)
3147 goto nla_put_failure;
3149 if (rt->rt6i_flags & RTF_GATEWAY) {
3150 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3151 goto nla_put_failure;
3155 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3156 goto nla_put_failure;
3157 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3158 goto nla_put_failure;
3160 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3162 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3163 goto nla_put_failure;
3165 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3166 goto nla_put_failure;
3168 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3170 nlmsg_end(skb, nlh);
3174 nlmsg_cancel(skb, nlh);
3178 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3180 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3183 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3184 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3185 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3189 return rt6_fill_node(arg->net,
3190 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3191 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3192 prefix, 0, NLM_F_MULTI);
3195 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3197 struct net *net = sock_net(in_skb->sk);
3198 struct nlattr *tb[RTA_MAX+1];
3199 struct rt6_info *rt;
3200 struct sk_buff *skb;
3203 int err, iif = 0, oif = 0;
3205 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3210 memset(&fl6, 0, sizeof(fl6));
3213 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3216 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3220 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3223 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3227 iif = nla_get_u32(tb[RTA_IIF]);
3230 oif = nla_get_u32(tb[RTA_OIF]);
3233 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3236 struct net_device *dev;
3239 dev = __dev_get_by_index(net, iif);
3245 fl6.flowi6_iif = iif;
3247 if (!ipv6_addr_any(&fl6.saddr))
3248 flags |= RT6_LOOKUP_F_HAS_SADDR;
3250 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3253 fl6.flowi6_oif = oif;
3255 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3258 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3265 /* Reserve room for dummy headers, this skb can pass
3266 through good chunk of routing engine.
3268 skb_reset_mac_header(skb);
3269 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3271 skb_dst_set(skb, &rt->dst);
3273 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3274 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3275 nlh->nlmsg_seq, 0, 0, 0);
3281 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3286 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3287 unsigned int nlm_flags)
3289 struct sk_buff *skb;
3290 struct net *net = info->nl_net;
3295 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3297 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3301 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3302 event, info->portid, seq, 0, 0, nlm_flags);
3304 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3305 WARN_ON(err == -EMSGSIZE);
3309 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3310 info->nlh, gfp_any());
3314 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3317 static int ip6_route_dev_notify(struct notifier_block *this,
3318 unsigned long event, void *ptr)
3320 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3321 struct net *net = dev_net(dev);
3323 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3324 net->ipv6.ip6_null_entry->dst.dev = dev;
3325 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3326 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3327 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3328 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3329 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3330 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3341 #ifdef CONFIG_PROC_FS
3343 static const struct file_operations ipv6_route_proc_fops = {
3344 .owner = THIS_MODULE,
3345 .open = ipv6_route_open,
3347 .llseek = seq_lseek,
3348 .release = seq_release_net,
3351 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3353 struct net *net = (struct net *)seq->private;
3354 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3355 net->ipv6.rt6_stats->fib_nodes,
3356 net->ipv6.rt6_stats->fib_route_nodes,
3357 net->ipv6.rt6_stats->fib_rt_alloc,
3358 net->ipv6.rt6_stats->fib_rt_entries,
3359 net->ipv6.rt6_stats->fib_rt_cache,
3360 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3361 net->ipv6.rt6_stats->fib_discarded_routes);
3366 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3368 return single_open_net(inode, file, rt6_stats_seq_show);
3371 static const struct file_operations rt6_stats_seq_fops = {
3372 .owner = THIS_MODULE,
3373 .open = rt6_stats_seq_open,
3375 .llseek = seq_lseek,
3376 .release = single_release_net,
3378 #endif /* CONFIG_PROC_FS */
3380 #ifdef CONFIG_SYSCTL
3383 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3384 void __user *buffer, size_t *lenp, loff_t *ppos)
3391 net = (struct net *)ctl->extra1;
3392 delay = net->ipv6.sysctl.flush_delay;
3393 proc_dointvec(ctl, write, buffer, lenp, ppos);
3394 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3398 struct ctl_table ipv6_route_table_template[] = {
3400 .procname = "flush",
3401 .data = &init_net.ipv6.sysctl.flush_delay,
3402 .maxlen = sizeof(int),
3404 .proc_handler = ipv6_sysctl_rtcache_flush
3407 .procname = "gc_thresh",
3408 .data = &ip6_dst_ops_template.gc_thresh,
3409 .maxlen = sizeof(int),
3411 .proc_handler = proc_dointvec,
3414 .procname = "max_size",
3415 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3416 .maxlen = sizeof(int),
3418 .proc_handler = proc_dointvec,
3421 .procname = "gc_min_interval",
3422 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3423 .maxlen = sizeof(int),
3425 .proc_handler = proc_dointvec_jiffies,
3428 .procname = "gc_timeout",
3429 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3430 .maxlen = sizeof(int),
3432 .proc_handler = proc_dointvec_jiffies,
3435 .procname = "gc_interval",
3436 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3437 .maxlen = sizeof(int),
3439 .proc_handler = proc_dointvec_jiffies,
3442 .procname = "gc_elasticity",
3443 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3444 .maxlen = sizeof(int),
3446 .proc_handler = proc_dointvec,
3449 .procname = "mtu_expires",
3450 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3451 .maxlen = sizeof(int),
3453 .proc_handler = proc_dointvec_jiffies,
3456 .procname = "min_adv_mss",
3457 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3458 .maxlen = sizeof(int),
3460 .proc_handler = proc_dointvec,
3463 .procname = "gc_min_interval_ms",
3464 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3465 .maxlen = sizeof(int),
3467 .proc_handler = proc_dointvec_ms_jiffies,
3472 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3474 struct ctl_table *table;
3476 table = kmemdup(ipv6_route_table_template,
3477 sizeof(ipv6_route_table_template),
3481 table[0].data = &net->ipv6.sysctl.flush_delay;
3482 table[0].extra1 = net;
3483 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3484 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3485 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3486 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3487 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3488 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3489 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3490 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3491 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3493 /* Don't export sysctls to unprivileged users */
3494 if (net->user_ns != &init_user_ns)
3495 table[0].procname = NULL;
3502 static int __net_init ip6_route_net_init(struct net *net)
3506 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3507 sizeof(net->ipv6.ip6_dst_ops));
3509 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3510 goto out_ip6_dst_ops;
3512 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3513 sizeof(*net->ipv6.ip6_null_entry),
3515 if (!net->ipv6.ip6_null_entry)
3516 goto out_ip6_dst_entries;
3517 net->ipv6.ip6_null_entry->dst.path =
3518 (struct dst_entry *)net->ipv6.ip6_null_entry;
3519 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3520 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3521 ip6_template_metrics, true);
3523 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3524 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3525 sizeof(*net->ipv6.ip6_prohibit_entry),
3527 if (!net->ipv6.ip6_prohibit_entry)
3528 goto out_ip6_null_entry;
3529 net->ipv6.ip6_prohibit_entry->dst.path =
3530 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3531 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3532 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3533 ip6_template_metrics, true);
3535 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3536 sizeof(*net->ipv6.ip6_blk_hole_entry),
3538 if (!net->ipv6.ip6_blk_hole_entry)
3539 goto out_ip6_prohibit_entry;
3540 net->ipv6.ip6_blk_hole_entry->dst.path =
3541 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3542 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3543 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3544 ip6_template_metrics, true);
3547 net->ipv6.sysctl.flush_delay = 0;
3548 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3549 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3550 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3551 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3552 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3553 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3554 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3556 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3562 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3563 out_ip6_prohibit_entry:
3564 kfree(net->ipv6.ip6_prohibit_entry);
3566 kfree(net->ipv6.ip6_null_entry);
3568 out_ip6_dst_entries:
3569 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3574 static void __net_exit ip6_route_net_exit(struct net *net)
3576 kfree(net->ipv6.ip6_null_entry);
3577 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3578 kfree(net->ipv6.ip6_prohibit_entry);
3579 kfree(net->ipv6.ip6_blk_hole_entry);
3581 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3584 static int __net_init ip6_route_net_init_late(struct net *net)
3586 #ifdef CONFIG_PROC_FS
3587 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3588 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3593 static void __net_exit ip6_route_net_exit_late(struct net *net)
3595 #ifdef CONFIG_PROC_FS
3596 remove_proc_entry("ipv6_route", net->proc_net);
3597 remove_proc_entry("rt6_stats", net->proc_net);
3601 static struct pernet_operations ip6_route_net_ops = {
3602 .init = ip6_route_net_init,
3603 .exit = ip6_route_net_exit,
3606 static int __net_init ipv6_inetpeer_init(struct net *net)
3608 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3612 inet_peer_base_init(bp);
3613 net->ipv6.peers = bp;
3617 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3619 struct inet_peer_base *bp = net->ipv6.peers;
3621 net->ipv6.peers = NULL;
3622 inetpeer_invalidate_tree(bp);
3626 static struct pernet_operations ipv6_inetpeer_ops = {
3627 .init = ipv6_inetpeer_init,
3628 .exit = ipv6_inetpeer_exit,
3631 static struct pernet_operations ip6_route_net_late_ops = {
3632 .init = ip6_route_net_init_late,
3633 .exit = ip6_route_net_exit_late,
3636 static struct notifier_block ip6_route_dev_notifier = {
3637 .notifier_call = ip6_route_dev_notify,
3641 int __init ip6_route_init(void)
3647 ip6_dst_ops_template.kmem_cachep =
3648 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3649 SLAB_HWCACHE_ALIGN, NULL);
3650 if (!ip6_dst_ops_template.kmem_cachep)
3653 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3655 goto out_kmem_cache;
3657 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3659 goto out_dst_entries;
3661 ret = register_pernet_subsys(&ip6_route_net_ops);
3663 goto out_register_inetpeer;
3665 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3667 /* Registering of the loopback is done before this portion of code,
3668 * the loopback reference in rt6_info will not be taken, do it
3669 * manually for init_net */
3670 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3671 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3672 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3673 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3674 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3675 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3676 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3680 goto out_register_subsys;
3686 ret = fib6_rules_init();
3690 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3692 goto fib6_rules_init;
3695 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3696 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3697 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3698 goto out_register_late_subsys;
3700 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3702 goto out_register_late_subsys;
3704 for_each_possible_cpu(cpu) {
3705 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3707 INIT_LIST_HEAD(&ul->head);
3708 spin_lock_init(&ul->lock);
3714 out_register_late_subsys:
3715 unregister_pernet_subsys(&ip6_route_net_late_ops);
3717 fib6_rules_cleanup();
3722 out_register_subsys:
3723 unregister_pernet_subsys(&ip6_route_net_ops);
3724 out_register_inetpeer:
3725 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3727 dst_entries_destroy(&ip6_dst_blackhole_ops);
3729 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3733 void ip6_route_cleanup(void)
3735 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3736 unregister_pernet_subsys(&ip6_route_net_late_ops);
3737 fib6_rules_cleanup();
3740 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3741 unregister_pernet_subsys(&ip6_route_net_ops);
3742 dst_entries_destroy(&ip6_dst_blackhole_ops);
3743 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);