2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
57 #include <net/dst_metadata.h>
59 #include <net/netevent.h>
60 #include <net/netlink.h>
61 #include <net/nexthop.h>
62 #include <net/lwtunnel.h>
63 #include <net/ip_tunnels.h>
64 #include <net/l3mdev.h>
65 #include <trace/events/fib6.h>
67 #include <asm/uaccess.h>
70 #include <linux/sysctl.h>
74 RT6_NUD_FAIL_HARD = -3,
75 RT6_NUD_FAIL_PROBE = -2,
76 RT6_NUD_FAIL_DO_RR = -1,
80 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
81 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
82 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
83 static unsigned int ip6_mtu(const struct dst_entry *dst);
84 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
85 static void ip6_dst_destroy(struct dst_entry *);
86 static void ip6_dst_ifdown(struct dst_entry *,
87 struct net_device *dev, int how);
88 static int ip6_dst_gc(struct dst_ops *ops);
90 static int ip6_pkt_discard(struct sk_buff *skb);
91 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
92 static int ip6_pkt_prohibit(struct sk_buff *skb);
93 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
94 static void ip6_link_failure(struct sk_buff *skb);
95 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
96 struct sk_buff *skb, u32 mtu);
97 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
99 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
100 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
102 #ifdef CONFIG_IPV6_ROUTE_INFO
103 static struct rt6_info *rt6_add_route_info(struct net *net,
104 const struct in6_addr *prefix, int prefixlen,
105 const struct in6_addr *gwaddr,
106 struct net_device *dev,
108 static struct rt6_info *rt6_get_route_info(struct net *net,
109 const struct in6_addr *prefix, int prefixlen,
110 const struct in6_addr *gwaddr,
111 struct net_device *dev);
114 struct uncached_list {
116 struct list_head head;
119 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
121 static void rt6_uncached_list_add(struct rt6_info *rt)
123 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
125 rt->dst.flags |= DST_NOCACHE;
126 rt->rt6i_uncached_list = ul;
128 spin_lock_bh(&ul->lock);
129 list_add_tail(&rt->rt6i_uncached, &ul->head);
130 spin_unlock_bh(&ul->lock);
133 static void rt6_uncached_list_del(struct rt6_info *rt)
135 if (!list_empty(&rt->rt6i_uncached)) {
136 struct uncached_list *ul = rt->rt6i_uncached_list;
138 spin_lock_bh(&ul->lock);
139 list_del(&rt->rt6i_uncached);
140 spin_unlock_bh(&ul->lock);
144 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
146 struct net_device *loopback_dev = net->loopback_dev;
149 if (dev == loopback_dev)
152 for_each_possible_cpu(cpu) {
153 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
156 spin_lock_bh(&ul->lock);
157 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
158 struct inet6_dev *rt_idev = rt->rt6i_idev;
159 struct net_device *rt_dev = rt->dst.dev;
161 if (rt_idev->dev == dev) {
162 rt->rt6i_idev = in6_dev_get(loopback_dev);
163 in6_dev_put(rt_idev);
167 rt->dst.dev = loopback_dev;
168 dev_hold(rt->dst.dev);
172 spin_unlock_bh(&ul->lock);
176 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
178 return dst_metrics_write_ptr(rt->dst.from);
181 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
183 struct rt6_info *rt = (struct rt6_info *)dst;
185 if (rt->rt6i_flags & RTF_PCPU)
186 return rt6_pcpu_cow_metrics(rt);
187 else if (rt->rt6i_flags & RTF_CACHE)
190 return dst_cow_metrics_generic(dst, old);
193 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
197 struct in6_addr *p = &rt->rt6i_gateway;
199 if (!ipv6_addr_any(p))
200 return (const void *) p;
202 return &ipv6_hdr(skb)->daddr;
206 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
210 struct rt6_info *rt = (struct rt6_info *) dst;
213 daddr = choose_neigh_daddr(rt, skb, daddr);
214 n = __ipv6_neigh_lookup(dst->dev, daddr);
217 return neigh_create(&nd_tbl, daddr, dst->dev);
220 static struct dst_ops ip6_dst_ops_template = {
224 .check = ip6_dst_check,
225 .default_advmss = ip6_default_advmss,
227 .cow_metrics = ipv6_cow_metrics,
228 .destroy = ip6_dst_destroy,
229 .ifdown = ip6_dst_ifdown,
230 .negative_advice = ip6_negative_advice,
231 .link_failure = ip6_link_failure,
232 .update_pmtu = ip6_rt_update_pmtu,
233 .redirect = rt6_do_redirect,
234 .local_out = __ip6_local_out,
235 .neigh_lookup = ip6_neigh_lookup,
238 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
240 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
242 return mtu ? : dst->dev->mtu;
245 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
246 struct sk_buff *skb, u32 mtu)
250 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
255 static struct dst_ops ip6_dst_blackhole_ops = {
257 .destroy = ip6_dst_destroy,
258 .check = ip6_dst_check,
259 .mtu = ip6_blackhole_mtu,
260 .default_advmss = ip6_default_advmss,
261 .update_pmtu = ip6_rt_blackhole_update_pmtu,
262 .redirect = ip6_rt_blackhole_redirect,
263 .cow_metrics = dst_cow_metrics_generic,
264 .neigh_lookup = ip6_neigh_lookup,
267 static const u32 ip6_template_metrics[RTAX_MAX] = {
268 [RTAX_HOPLIMIT - 1] = 0,
271 static const struct rt6_info ip6_null_entry_template = {
273 .__refcnt = ATOMIC_INIT(1),
275 .obsolete = DST_OBSOLETE_FORCE_CHK,
276 .error = -ENETUNREACH,
277 .input = ip6_pkt_discard,
278 .output = ip6_pkt_discard_out,
280 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
281 .rt6i_protocol = RTPROT_KERNEL,
282 .rt6i_metric = ~(u32) 0,
283 .rt6i_ref = ATOMIC_INIT(1),
286 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
288 static const struct rt6_info ip6_prohibit_entry_template = {
290 .__refcnt = ATOMIC_INIT(1),
292 .obsolete = DST_OBSOLETE_FORCE_CHK,
294 .input = ip6_pkt_prohibit,
295 .output = ip6_pkt_prohibit_out,
297 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
298 .rt6i_protocol = RTPROT_KERNEL,
299 .rt6i_metric = ~(u32) 0,
300 .rt6i_ref = ATOMIC_INIT(1),
303 static const struct rt6_info ip6_blk_hole_entry_template = {
305 .__refcnt = ATOMIC_INIT(1),
307 .obsolete = DST_OBSOLETE_FORCE_CHK,
309 .input = dst_discard,
310 .output = dst_discard_out,
312 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
313 .rt6i_protocol = RTPROT_KERNEL,
314 .rt6i_metric = ~(u32) 0,
315 .rt6i_ref = ATOMIC_INIT(1),
320 static void rt6_info_init(struct rt6_info *rt)
322 struct dst_entry *dst = &rt->dst;
324 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
325 INIT_LIST_HEAD(&rt->rt6i_siblings);
326 INIT_LIST_HEAD(&rt->rt6i_uncached);
329 /* allocate dst with ip6_dst_ops */
330 static struct rt6_info *__ip6_dst_alloc(struct net *net,
331 struct net_device *dev,
334 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
335 0, DST_OBSOLETE_FORCE_CHK, flags);
343 struct rt6_info *ip6_dst_alloc(struct net *net,
344 struct net_device *dev,
347 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
350 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
354 for_each_possible_cpu(cpu) {
357 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
358 /* no one shares rt */
362 dst_destroy((struct dst_entry *)rt);
369 EXPORT_SYMBOL(ip6_dst_alloc);
371 static void ip6_dst_destroy(struct dst_entry *dst)
373 struct rt6_info *rt = (struct rt6_info *)dst;
374 struct dst_entry *from = dst->from;
375 struct inet6_dev *idev;
377 dst_destroy_metrics_generic(dst);
378 free_percpu(rt->rt6i_pcpu);
379 rt6_uncached_list_del(rt);
381 idev = rt->rt6i_idev;
383 rt->rt6i_idev = NULL;
391 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
394 struct rt6_info *rt = (struct rt6_info *)dst;
395 struct inet6_dev *idev = rt->rt6i_idev;
396 struct net_device *loopback_dev =
397 dev_net(dev)->loopback_dev;
399 if (dev != loopback_dev) {
400 if (idev && idev->dev == dev) {
401 struct inet6_dev *loopback_idev =
402 in6_dev_get(loopback_dev);
404 rt->rt6i_idev = loopback_idev;
411 static bool __rt6_check_expired(const struct rt6_info *rt)
413 if (rt->rt6i_flags & RTF_EXPIRES)
414 return time_after(jiffies, rt->dst.expires);
419 static bool rt6_check_expired(const struct rt6_info *rt)
421 if (rt->rt6i_flags & RTF_EXPIRES) {
422 if (time_after(jiffies, rt->dst.expires))
424 } else if (rt->dst.from) {
425 return rt6_check_expired((struct rt6_info *) rt->dst.from);
430 /* Multipath route selection:
431 * Hash based function using packet header and flowlabel.
432 * Adapted from fib_info_hashfn()
434 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
435 const struct flowi6 *fl6)
437 return get_hash_from_flowi6(fl6) % candidate_count;
440 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
441 struct flowi6 *fl6, int oif,
444 struct rt6_info *sibling, *next_sibling;
447 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
448 /* Don't change the route, if route_choosen == 0
449 * (siblings does not include ourself)
452 list_for_each_entry_safe(sibling, next_sibling,
453 &match->rt6i_siblings, rt6i_siblings) {
455 if (route_choosen == 0) {
456 if (rt6_score_route(sibling, oif, strict) < 0)
466 * Route lookup. Any table->tb6_lock is implied.
469 static inline struct rt6_info *rt6_device_match(struct net *net,
471 const struct in6_addr *saddr,
475 struct rt6_info *local = NULL;
476 struct rt6_info *sprt;
478 if (!oif && ipv6_addr_any(saddr))
481 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
482 struct net_device *dev = sprt->dst.dev;
485 if (dev->ifindex == oif)
487 if (dev->flags & IFF_LOOPBACK) {
488 if (!sprt->rt6i_idev ||
489 sprt->rt6i_idev->dev->ifindex != oif) {
490 if (flags & RT6_LOOKUP_F_IFACE)
493 local->rt6i_idev->dev->ifindex == oif)
499 if (ipv6_chk_addr(net, saddr, dev,
500 flags & RT6_LOOKUP_F_IFACE))
509 if (flags & RT6_LOOKUP_F_IFACE)
510 return net->ipv6.ip6_null_entry;
516 #ifdef CONFIG_IPV6_ROUTER_PREF
517 struct __rt6_probe_work {
518 struct work_struct work;
519 struct in6_addr target;
520 struct net_device *dev;
523 static void rt6_probe_deferred(struct work_struct *w)
525 struct in6_addr mcaddr;
526 struct __rt6_probe_work *work =
527 container_of(w, struct __rt6_probe_work, work);
529 addrconf_addr_solict_mult(&work->target, &mcaddr);
530 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
535 static void rt6_probe(struct rt6_info *rt)
537 struct __rt6_probe_work *work;
538 struct neighbour *neigh;
540 * Okay, this does not seem to be appropriate
541 * for now, however, we need to check if it
542 * is really so; aka Router Reachability Probing.
544 * Router Reachability Probe MUST be rate-limited
545 * to no more than one per minute.
547 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
550 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
552 if (neigh->nud_state & NUD_VALID)
556 write_lock(&neigh->lock);
557 if (!(neigh->nud_state & NUD_VALID) &&
560 rt->rt6i_idev->cnf.rtr_probe_interval)) {
561 work = kmalloc(sizeof(*work), GFP_ATOMIC);
563 __neigh_set_probe_once(neigh);
565 write_unlock(&neigh->lock);
567 work = kmalloc(sizeof(*work), GFP_ATOMIC);
571 INIT_WORK(&work->work, rt6_probe_deferred);
572 work->target = rt->rt6i_gateway;
573 dev_hold(rt->dst.dev);
574 work->dev = rt->dst.dev;
575 schedule_work(&work->work);
579 rcu_read_unlock_bh();
582 static inline void rt6_probe(struct rt6_info *rt)
588 * Default Router Selection (RFC 2461 6.3.6)
590 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
592 struct net_device *dev = rt->dst.dev;
593 if (!oif || dev->ifindex == oif)
595 if ((dev->flags & IFF_LOOPBACK) &&
596 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
601 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
603 struct neighbour *neigh;
604 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
606 if (rt->rt6i_flags & RTF_NONEXTHOP ||
607 !(rt->rt6i_flags & RTF_GATEWAY))
608 return RT6_NUD_SUCCEED;
611 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
613 read_lock(&neigh->lock);
614 if (neigh->nud_state & NUD_VALID)
615 ret = RT6_NUD_SUCCEED;
616 #ifdef CONFIG_IPV6_ROUTER_PREF
617 else if (!(neigh->nud_state & NUD_FAILED))
618 ret = RT6_NUD_SUCCEED;
620 ret = RT6_NUD_FAIL_PROBE;
622 read_unlock(&neigh->lock);
624 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
625 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
627 rcu_read_unlock_bh();
632 static int rt6_score_route(struct rt6_info *rt, int oif,
637 m = rt6_check_dev(rt, oif);
638 if (!m && (strict & RT6_LOOKUP_F_IFACE))
639 return RT6_NUD_FAIL_HARD;
640 #ifdef CONFIG_IPV6_ROUTER_PREF
641 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
643 if (strict & RT6_LOOKUP_F_REACHABLE) {
644 int n = rt6_check_neigh(rt);
651 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
652 int *mpri, struct rt6_info *match,
656 bool match_do_rr = false;
657 struct inet6_dev *idev = rt->rt6i_idev;
658 struct net_device *dev = rt->dst.dev;
660 if (dev && !netif_carrier_ok(dev) &&
661 idev->cnf.ignore_routes_with_linkdown &&
662 !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
665 if (rt6_check_expired(rt))
668 m = rt6_score_route(rt, oif, strict);
669 if (m == RT6_NUD_FAIL_DO_RR) {
671 m = 0; /* lowest valid score */
672 } else if (m == RT6_NUD_FAIL_HARD) {
676 if (strict & RT6_LOOKUP_F_REACHABLE)
679 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
681 *do_rr = match_do_rr;
689 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
690 struct rt6_info *rr_head,
691 u32 metric, int oif, int strict,
694 struct rt6_info *rt, *match, *cont;
699 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
700 if (rt->rt6i_metric != metric) {
705 match = find_match(rt, oif, strict, &mpri, match, do_rr);
708 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
709 if (rt->rt6i_metric != metric) {
714 match = find_match(rt, oif, strict, &mpri, match, do_rr);
720 for (rt = cont; rt; rt = rt->dst.rt6_next)
721 match = find_match(rt, oif, strict, &mpri, match, do_rr);
726 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
728 struct rt6_info *match, *rt0;
734 fn->rr_ptr = rt0 = fn->leaf;
736 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
740 struct rt6_info *next = rt0->dst.rt6_next;
742 /* no entries matched; do round-robin */
743 if (!next || next->rt6i_metric != rt0->rt6i_metric)
750 net = dev_net(rt0->dst.dev);
751 return match ? match : net->ipv6.ip6_null_entry;
754 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
756 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
759 #ifdef CONFIG_IPV6_ROUTE_INFO
760 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
761 const struct in6_addr *gwaddr)
763 struct net *net = dev_net(dev);
764 struct route_info *rinfo = (struct route_info *) opt;
765 struct in6_addr prefix_buf, *prefix;
767 unsigned long lifetime;
770 if (len < sizeof(struct route_info)) {
774 /* Sanity check for prefix_len and length */
775 if (rinfo->length > 3) {
777 } else if (rinfo->prefix_len > 128) {
779 } else if (rinfo->prefix_len > 64) {
780 if (rinfo->length < 2) {
783 } else if (rinfo->prefix_len > 0) {
784 if (rinfo->length < 1) {
789 pref = rinfo->route_pref;
790 if (pref == ICMPV6_ROUTER_PREF_INVALID)
793 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
795 if (rinfo->length == 3)
796 prefix = (struct in6_addr *)rinfo->prefix;
798 /* this function is safe */
799 ipv6_addr_prefix(&prefix_buf,
800 (struct in6_addr *)rinfo->prefix,
802 prefix = &prefix_buf;
805 if (rinfo->prefix_len == 0)
806 rt = rt6_get_dflt_router(gwaddr, dev);
808 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
811 if (rt && !lifetime) {
817 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
820 rt->rt6i_flags = RTF_ROUTEINFO |
821 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
824 if (!addrconf_finite_timeout(lifetime))
825 rt6_clean_expires(rt);
827 rt6_set_expires(rt, jiffies + HZ * lifetime);
835 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
836 struct in6_addr *saddr)
838 struct fib6_node *pn;
840 if (fn->fn_flags & RTN_TL_ROOT)
843 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
844 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
847 if (fn->fn_flags & RTN_RTINFO)
852 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
853 struct fib6_table *table,
854 struct flowi6 *fl6, int flags)
856 struct fib6_node *fn;
859 read_lock_bh(&table->tb6_lock);
860 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
863 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
864 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
865 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
866 if (rt == net->ipv6.ip6_null_entry) {
867 fn = fib6_backtrack(fn, &fl6->saddr);
871 dst_use(&rt->dst, jiffies);
872 read_unlock_bh(&table->tb6_lock);
874 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
880 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
883 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
885 EXPORT_SYMBOL_GPL(ip6_route_lookup);
887 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
888 const struct in6_addr *saddr, int oif, int strict)
890 struct flowi6 fl6 = {
894 struct dst_entry *dst;
895 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
898 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
899 flags |= RT6_LOOKUP_F_HAS_SADDR;
902 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
904 return (struct rt6_info *) dst;
910 EXPORT_SYMBOL(rt6_lookup);
912 /* ip6_ins_rt is called with FREE table->tb6_lock.
913 It takes new route entry, the addition fails by any reason the
914 route is freed. In any case, if caller does not hold it, it may
918 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
919 struct mx6_config *mxc)
922 struct fib6_table *table;
924 table = rt->rt6i_table;
925 write_lock_bh(&table->tb6_lock);
926 err = fib6_add(&table->tb6_root, rt, info, mxc);
927 write_unlock_bh(&table->tb6_lock);
932 int ip6_ins_rt(struct rt6_info *rt)
934 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
935 struct mx6_config mxc = { .mx = NULL, };
937 return __ip6_ins_rt(rt, &info, &mxc);
940 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
941 const struct in6_addr *daddr,
942 const struct in6_addr *saddr)
950 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
951 ort = (struct rt6_info *)ort->dst.from;
953 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
958 ip6_rt_copy_init(rt, ort);
959 rt->rt6i_flags |= RTF_CACHE;
961 rt->dst.flags |= DST_HOST;
962 rt->rt6i_dst.addr = *daddr;
963 rt->rt6i_dst.plen = 128;
965 if (!rt6_is_gw_or_nonexthop(ort)) {
966 if (ort->rt6i_dst.plen != 128 &&
967 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
968 rt->rt6i_flags |= RTF_ANYCAST;
969 #ifdef CONFIG_IPV6_SUBTREES
970 if (rt->rt6i_src.plen && saddr) {
971 rt->rt6i_src.addr = *saddr;
972 rt->rt6i_src.plen = 128;
980 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
982 struct rt6_info *pcpu_rt;
984 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
985 rt->dst.dev, rt->dst.flags);
989 ip6_rt_copy_init(pcpu_rt, rt);
990 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
991 pcpu_rt->rt6i_flags |= RTF_PCPU;
995 /* It should be called with read_lock_bh(&tb6_lock) acquired */
996 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
998 struct rt6_info *pcpu_rt, **p;
1000 p = this_cpu_ptr(rt->rt6i_pcpu);
1004 dst_hold(&pcpu_rt->dst);
1005 rt6_dst_from_metrics_check(pcpu_rt);
1010 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1012 struct fib6_table *table = rt->rt6i_table;
1013 struct rt6_info *pcpu_rt, *prev, **p;
1015 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1017 struct net *net = dev_net(rt->dst.dev);
1019 dst_hold(&net->ipv6.ip6_null_entry->dst);
1020 return net->ipv6.ip6_null_entry;
1023 read_lock_bh(&table->tb6_lock);
1024 if (rt->rt6i_pcpu) {
1025 p = this_cpu_ptr(rt->rt6i_pcpu);
1026 prev = cmpxchg(p, NULL, pcpu_rt);
1028 /* If someone did it before us, return prev instead */
1029 dst_destroy(&pcpu_rt->dst);
1033 /* rt has been removed from the fib6 tree
1034 * before we have a chance to acquire the read_lock.
1035 * In this case, don't brother to create a pcpu rt
1036 * since rt is going away anyway. The next
1037 * dst_check() will trigger a re-lookup.
1039 dst_destroy(&pcpu_rt->dst);
1042 dst_hold(&pcpu_rt->dst);
1043 rt6_dst_from_metrics_check(pcpu_rt);
1044 read_unlock_bh(&table->tb6_lock);
1048 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1049 int oif, struct flowi6 *fl6, int flags)
1051 struct fib6_node *fn, *saved_fn;
1052 struct rt6_info *rt;
1055 strict |= flags & RT6_LOOKUP_F_IFACE;
1056 strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1057 if (net->ipv6.devconf_all->forwarding == 0)
1058 strict |= RT6_LOOKUP_F_REACHABLE;
1060 read_lock_bh(&table->tb6_lock);
1062 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1065 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1069 rt = rt6_select(fn, oif, strict);
1070 if (rt->rt6i_nsiblings)
1071 rt = rt6_multipath_select(rt, fl6, oif, strict);
1072 if (rt == net->ipv6.ip6_null_entry) {
1073 fn = fib6_backtrack(fn, &fl6->saddr);
1075 goto redo_rt6_select;
1076 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1077 /* also consider unreachable route */
1078 strict &= ~RT6_LOOKUP_F_REACHABLE;
1080 goto redo_rt6_select;
1085 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1086 dst_use(&rt->dst, jiffies);
1087 read_unlock_bh(&table->tb6_lock);
1089 rt6_dst_from_metrics_check(rt);
1091 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1093 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1094 !(rt->rt6i_flags & RTF_GATEWAY))) {
1095 /* Create a RTF_CACHE clone which will not be
1096 * owned by the fib6 tree. It is for the special case where
1097 * the daddr in the skb during the neighbor look-up is different
1098 * from the fl6->daddr used to look-up route here.
1101 struct rt6_info *uncached_rt;
1103 dst_use(&rt->dst, jiffies);
1104 read_unlock_bh(&table->tb6_lock);
1106 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1107 dst_release(&rt->dst);
1110 rt6_uncached_list_add(uncached_rt);
1112 uncached_rt = net->ipv6.ip6_null_entry;
1114 dst_hold(&uncached_rt->dst);
1116 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1120 /* Get a percpu copy */
1122 struct rt6_info *pcpu_rt;
1124 rt->dst.lastuse = jiffies;
1126 pcpu_rt = rt6_get_pcpu_route(rt);
1129 read_unlock_bh(&table->tb6_lock);
1131 /* We have to do the read_unlock first
1132 * because rt6_make_pcpu_route() may trigger
1133 * ip6_dst_gc() which will take the write_lock.
1136 read_unlock_bh(&table->tb6_lock);
1137 pcpu_rt = rt6_make_pcpu_route(rt);
1138 dst_release(&rt->dst);
1141 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1146 EXPORT_SYMBOL_GPL(ip6_pol_route);
1148 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1149 struct flowi6 *fl6, int flags)
1151 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1154 struct dst_entry *ip6_route_input_lookup(struct net *net,
1155 struct net_device *dev,
1156 struct flowi6 *fl6, int flags)
1158 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1159 flags |= RT6_LOOKUP_F_IFACE;
1161 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1163 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1165 void ip6_route_input(struct sk_buff *skb)
1167 const struct ipv6hdr *iph = ipv6_hdr(skb);
1168 struct net *net = dev_net(skb->dev);
1169 int flags = RT6_LOOKUP_F_HAS_SADDR;
1170 struct ip_tunnel_info *tun_info;
1171 struct flowi6 fl6 = {
1172 .flowi6_iif = skb->dev->ifindex,
1173 .daddr = iph->daddr,
1174 .saddr = iph->saddr,
1175 .flowlabel = ip6_flowinfo(iph),
1176 .flowi6_mark = skb->mark,
1177 .flowi6_proto = iph->nexthdr,
1180 tun_info = skb_tunnel_info(skb);
1181 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1182 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1184 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1187 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1188 struct flowi6 *fl6, int flags)
1190 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1193 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1194 struct flowi6 *fl6, int flags)
1198 if (rt6_need_strict(&fl6->daddr)) {
1199 struct dst_entry *dst;
1201 dst = l3mdev_link_scope_lookup(net, fl6);
1206 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1208 any_src = ipv6_addr_any(&fl6->saddr);
1209 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1210 (fl6->flowi6_oif && any_src))
1211 flags |= RT6_LOOKUP_F_IFACE;
1214 flags |= RT6_LOOKUP_F_HAS_SADDR;
1216 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1218 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1220 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1222 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1224 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1225 struct dst_entry *new = NULL;
1227 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1233 new->input = dst_discard;
1234 new->output = dst_discard_out;
1236 dst_copy_metrics(new, &ort->dst);
1237 rt->rt6i_idev = ort->rt6i_idev;
1239 in6_dev_hold(rt->rt6i_idev);
1241 rt->rt6i_gateway = ort->rt6i_gateway;
1242 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1243 rt->rt6i_metric = 0;
1245 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1246 #ifdef CONFIG_IPV6_SUBTREES
1247 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1253 dst_release(dst_orig);
1254 return new ? new : ERR_PTR(-ENOMEM);
1258 * Destination cache support functions
1261 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1264 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1265 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1268 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1270 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1273 if (rt6_check_expired(rt))
1279 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1281 if (!__rt6_check_expired(rt) &&
1282 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1283 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1289 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1291 struct rt6_info *rt;
1293 rt = (struct rt6_info *) dst;
1295 /* All IPV6 dsts are created with ->obsolete set to the value
1296 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1297 * into this function always.
1300 rt6_dst_from_metrics_check(rt);
1302 if (rt->rt6i_flags & RTF_PCPU ||
1303 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1304 return rt6_dst_from_check(rt, cookie);
1306 return rt6_check(rt, cookie);
1309 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1311 struct rt6_info *rt = (struct rt6_info *) dst;
1314 if (rt->rt6i_flags & RTF_CACHE) {
1315 if (rt6_check_expired(rt)) {
1327 static void ip6_link_failure(struct sk_buff *skb)
1329 struct rt6_info *rt;
1331 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1333 rt = (struct rt6_info *) skb_dst(skb);
1335 if (rt->rt6i_flags & RTF_CACHE) {
1338 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1339 rt->rt6i_node->fn_sernum = -1;
1344 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1346 struct net *net = dev_net(rt->dst.dev);
1348 rt->rt6i_flags |= RTF_MODIFIED;
1349 rt->rt6i_pmtu = mtu;
1350 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1353 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1355 return !(rt->rt6i_flags & RTF_CACHE) &&
1356 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1359 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1360 const struct ipv6hdr *iph, u32 mtu)
1362 struct rt6_info *rt6 = (struct rt6_info *)dst;
1364 if (rt6->rt6i_flags & RTF_LOCAL)
1367 if (dst_metric_locked(dst, RTAX_MTU))
1371 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1372 if (mtu >= dst_mtu(dst))
1375 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1376 rt6_do_update_pmtu(rt6, mtu);
1378 const struct in6_addr *daddr, *saddr;
1379 struct rt6_info *nrt6;
1382 daddr = &iph->daddr;
1383 saddr = &iph->saddr;
1385 daddr = &sk->sk_v6_daddr;
1386 saddr = &inet6_sk(sk)->saddr;
1390 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1392 rt6_do_update_pmtu(nrt6, mtu);
1394 /* ip6_ins_rt(nrt6) will bump the
1395 * rt6->rt6i_node->fn_sernum
1396 * which will fail the next rt6_check() and
1397 * invalidate the sk->sk_dst_cache.
1404 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1405 struct sk_buff *skb, u32 mtu)
1407 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1410 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1413 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1414 struct dst_entry *dst;
1417 memset(&fl6, 0, sizeof(fl6));
1418 fl6.flowi6_oif = oif;
1419 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1420 fl6.daddr = iph->daddr;
1421 fl6.saddr = iph->saddr;
1422 fl6.flowlabel = ip6_flowinfo(iph);
1424 dst = ip6_route_output(net, NULL, &fl6);
1426 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1429 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1431 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1433 struct dst_entry *dst;
1435 ip6_update_pmtu(skb, sock_net(sk), mtu,
1436 sk->sk_bound_dev_if, sk->sk_mark);
1438 dst = __sk_dst_get(sk);
1439 if (!dst || !dst->obsolete ||
1440 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1444 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1445 ip6_datagram_dst_update(sk, false);
1448 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1450 /* Handle redirects */
1451 struct ip6rd_flowi {
1453 struct in6_addr gateway;
1456 static struct rt6_info *__ip6_route_redirect(struct net *net,
1457 struct fib6_table *table,
1461 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1462 struct rt6_info *rt;
1463 struct fib6_node *fn;
1465 /* Get the "current" route for this destination and
1466 * check if the redirect has come from approriate router.
1468 * RFC 4861 specifies that redirects should only be
1469 * accepted if they come from the nexthop to the target.
1470 * Due to the way the routes are chosen, this notion
1471 * is a bit fuzzy and one might need to check all possible
1475 read_lock_bh(&table->tb6_lock);
1476 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1478 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1479 if (rt6_check_expired(rt))
1483 if (!(rt->rt6i_flags & RTF_GATEWAY))
1485 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1487 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1493 rt = net->ipv6.ip6_null_entry;
1494 else if (rt->dst.error) {
1495 rt = net->ipv6.ip6_null_entry;
1499 if (rt == net->ipv6.ip6_null_entry) {
1500 fn = fib6_backtrack(fn, &fl6->saddr);
1508 read_unlock_bh(&table->tb6_lock);
1510 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1514 static struct dst_entry *ip6_route_redirect(struct net *net,
1515 const struct flowi6 *fl6,
1516 const struct in6_addr *gateway)
1518 int flags = RT6_LOOKUP_F_HAS_SADDR;
1519 struct ip6rd_flowi rdfl;
1522 rdfl.gateway = *gateway;
1524 return fib6_rule_lookup(net, &rdfl.fl6,
1525 flags, __ip6_route_redirect);
1528 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1530 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1531 struct dst_entry *dst;
1534 memset(&fl6, 0, sizeof(fl6));
1535 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1536 fl6.flowi6_oif = oif;
1537 fl6.flowi6_mark = mark;
1538 fl6.daddr = iph->daddr;
1539 fl6.saddr = iph->saddr;
1540 fl6.flowlabel = ip6_flowinfo(iph);
1542 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1543 rt6_do_redirect(dst, NULL, skb);
1546 EXPORT_SYMBOL_GPL(ip6_redirect);
1548 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1551 const struct ipv6hdr *iph = ipv6_hdr(skb);
1552 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1553 struct dst_entry *dst;
1556 memset(&fl6, 0, sizeof(fl6));
1557 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1558 fl6.flowi6_oif = oif;
1559 fl6.flowi6_mark = mark;
1560 fl6.daddr = msg->dest;
1561 fl6.saddr = iph->daddr;
1563 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1564 rt6_do_redirect(dst, NULL, skb);
1568 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1570 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1572 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1574 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1576 struct net_device *dev = dst->dev;
1577 unsigned int mtu = dst_mtu(dst);
1578 struct net *net = dev_net(dev);
1580 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1582 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1583 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1586 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1587 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1588 * IPV6_MAXPLEN is also valid and means: "any MSS,
1589 * rely only on pmtu discovery"
1591 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1596 static unsigned int ip6_mtu(const struct dst_entry *dst)
1598 const struct rt6_info *rt = (const struct rt6_info *)dst;
1599 unsigned int mtu = rt->rt6i_pmtu;
1600 struct inet6_dev *idev;
1605 mtu = dst_metric_raw(dst, RTAX_MTU);
1612 idev = __in6_dev_get(dst->dev);
1614 mtu = idev->cnf.mtu6;
1618 mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1620 return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1623 static struct dst_entry *icmp6_dst_gc_list;
1624 static DEFINE_SPINLOCK(icmp6_dst_lock);
1626 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1629 struct dst_entry *dst;
1630 struct rt6_info *rt;
1631 struct inet6_dev *idev = in6_dev_get(dev);
1632 struct net *net = dev_net(dev);
1634 if (unlikely(!idev))
1635 return ERR_PTR(-ENODEV);
1637 rt = ip6_dst_alloc(net, dev, 0);
1638 if (unlikely(!rt)) {
1640 dst = ERR_PTR(-ENOMEM);
1644 rt->dst.flags |= DST_HOST;
1645 rt->dst.output = ip6_output;
1646 atomic_set(&rt->dst.__refcnt, 1);
1647 rt->rt6i_gateway = fl6->daddr;
1648 rt->rt6i_dst.addr = fl6->daddr;
1649 rt->rt6i_dst.plen = 128;
1650 rt->rt6i_idev = idev;
1651 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1653 spin_lock_bh(&icmp6_dst_lock);
1654 rt->dst.next = icmp6_dst_gc_list;
1655 icmp6_dst_gc_list = &rt->dst;
1656 spin_unlock_bh(&icmp6_dst_lock);
1658 fib6_force_start_gc(net);
1660 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1666 int icmp6_dst_gc(void)
1668 struct dst_entry *dst, **pprev;
1671 spin_lock_bh(&icmp6_dst_lock);
1672 pprev = &icmp6_dst_gc_list;
1674 while ((dst = *pprev) != NULL) {
1675 if (!atomic_read(&dst->__refcnt)) {
1684 spin_unlock_bh(&icmp6_dst_lock);
1689 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1692 struct dst_entry *dst, **pprev;
1694 spin_lock_bh(&icmp6_dst_lock);
1695 pprev = &icmp6_dst_gc_list;
1696 while ((dst = *pprev) != NULL) {
1697 struct rt6_info *rt = (struct rt6_info *) dst;
1698 if (func(rt, arg)) {
1705 spin_unlock_bh(&icmp6_dst_lock);
1708 static int ip6_dst_gc(struct dst_ops *ops)
1710 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1711 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1712 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1713 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1714 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1715 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1718 entries = dst_entries_get_fast(ops);
1719 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1720 entries <= rt_max_size)
1723 net->ipv6.ip6_rt_gc_expire++;
1724 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1725 entries = dst_entries_get_slow(ops);
1726 if (entries < ops->gc_thresh)
1727 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1729 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1730 return entries > rt_max_size;
1733 static int ip6_convert_metrics(struct mx6_config *mxc,
1734 const struct fib6_config *cfg)
1736 bool ecn_ca = false;
1744 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1748 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1749 int type = nla_type(nla);
1754 if (unlikely(type > RTAX_MAX))
1757 if (type == RTAX_CC_ALGO) {
1758 char tmp[TCP_CA_NAME_MAX];
1760 nla_strlcpy(tmp, nla, sizeof(tmp));
1761 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1762 if (val == TCP_CA_UNSPEC)
1765 val = nla_get_u32(nla);
1767 if (type == RTAX_HOPLIMIT && val > 255)
1769 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1773 __set_bit(type - 1, mxc->mx_valid);
1777 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1778 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1788 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1789 struct fib6_config *cfg,
1790 const struct in6_addr *gw_addr)
1792 struct flowi6 fl6 = {
1793 .flowi6_oif = cfg->fc_ifindex,
1795 .saddr = cfg->fc_prefsrc,
1797 struct fib6_table *table;
1798 struct rt6_info *rt;
1799 int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
1801 table = fib6_get_table(net, cfg->fc_table);
1805 if (!ipv6_addr_any(&cfg->fc_prefsrc))
1806 flags |= RT6_LOOKUP_F_HAS_SADDR;
1808 rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1810 /* if table lookup failed, fall back to full lookup */
1811 if (rt == net->ipv6.ip6_null_entry) {
1819 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1821 struct net *net = cfg->fc_nlinfo.nl_net;
1822 struct rt6_info *rt = NULL;
1823 struct net_device *dev = NULL;
1824 struct inet6_dev *idev = NULL;
1825 struct fib6_table *table;
1829 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1831 #ifndef CONFIG_IPV6_SUBTREES
1832 if (cfg->fc_src_len)
1835 if (cfg->fc_ifindex) {
1837 dev = dev_get_by_index(net, cfg->fc_ifindex);
1840 idev = in6_dev_get(dev);
1845 if (cfg->fc_metric == 0)
1846 cfg->fc_metric = IP6_RT_PRIO_USER;
1849 if (cfg->fc_nlinfo.nlh &&
1850 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1851 table = fib6_get_table(net, cfg->fc_table);
1853 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1854 table = fib6_new_table(net, cfg->fc_table);
1857 table = fib6_new_table(net, cfg->fc_table);
1863 rt = ip6_dst_alloc(net, NULL,
1864 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1871 if (cfg->fc_flags & RTF_EXPIRES)
1872 rt6_set_expires(rt, jiffies +
1873 clock_t_to_jiffies(cfg->fc_expires));
1875 rt6_clean_expires(rt);
1877 if (cfg->fc_protocol == RTPROT_UNSPEC)
1878 cfg->fc_protocol = RTPROT_BOOT;
1879 rt->rt6i_protocol = cfg->fc_protocol;
1881 addr_type = ipv6_addr_type(&cfg->fc_dst);
1883 if (addr_type & IPV6_ADDR_MULTICAST)
1884 rt->dst.input = ip6_mc_input;
1885 else if (cfg->fc_flags & RTF_LOCAL)
1886 rt->dst.input = ip6_input;
1888 rt->dst.input = ip6_forward;
1890 rt->dst.output = ip6_output;
1892 if (cfg->fc_encap) {
1893 struct lwtunnel_state *lwtstate;
1895 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1896 cfg->fc_encap, AF_INET6, cfg,
1900 rt->dst.lwtstate = lwtstate_get(lwtstate);
1901 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1902 rt->dst.lwtstate->orig_output = rt->dst.output;
1903 rt->dst.output = lwtunnel_output;
1905 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1906 rt->dst.lwtstate->orig_input = rt->dst.input;
1907 rt->dst.input = lwtunnel_input;
1911 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1912 rt->rt6i_dst.plen = cfg->fc_dst_len;
1913 if (rt->rt6i_dst.plen == 128)
1914 rt->dst.flags |= DST_HOST;
1916 #ifdef CONFIG_IPV6_SUBTREES
1917 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1918 rt->rt6i_src.plen = cfg->fc_src_len;
1921 rt->rt6i_metric = cfg->fc_metric;
1923 /* We cannot add true routes via loopback here,
1924 they would result in kernel looping; promote them to reject routes
1926 if ((cfg->fc_flags & RTF_REJECT) ||
1927 (dev && (dev->flags & IFF_LOOPBACK) &&
1928 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1929 !(cfg->fc_flags & RTF_LOCAL))) {
1930 /* hold loopback dev/idev if we haven't done so. */
1931 if (dev != net->loopback_dev) {
1936 dev = net->loopback_dev;
1938 idev = in6_dev_get(dev);
1944 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1945 switch (cfg->fc_type) {
1947 rt->dst.error = -EINVAL;
1948 rt->dst.output = dst_discard_out;
1949 rt->dst.input = dst_discard;
1952 rt->dst.error = -EACCES;
1953 rt->dst.output = ip6_pkt_prohibit_out;
1954 rt->dst.input = ip6_pkt_prohibit;
1957 case RTN_UNREACHABLE:
1959 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1960 : (cfg->fc_type == RTN_UNREACHABLE)
1961 ? -EHOSTUNREACH : -ENETUNREACH;
1962 rt->dst.output = ip6_pkt_discard_out;
1963 rt->dst.input = ip6_pkt_discard;
1969 if (cfg->fc_flags & RTF_GATEWAY) {
1970 const struct in6_addr *gw_addr;
1973 gw_addr = &cfg->fc_gateway;
1974 gwa_type = ipv6_addr_type(gw_addr);
1976 /* if gw_addr is local we will fail to detect this in case
1977 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1978 * will return already-added prefix route via interface that
1979 * prefix route was assigned to, which might be non-loopback.
1982 if (ipv6_chk_addr_and_flags(net, gw_addr,
1983 gwa_type & IPV6_ADDR_LINKLOCAL ?
1987 rt->rt6i_gateway = *gw_addr;
1989 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1990 struct rt6_info *grt = NULL;
1992 /* IPv6 strictly inhibits using not link-local
1993 addresses as nexthop address.
1994 Otherwise, router will not able to send redirects.
1995 It is very good, but in some (rare!) circumstances
1996 (SIT, PtP, NBMA NOARP links) it is handy to allow
1997 some exceptions. --ANK
1999 if (!(gwa_type & IPV6_ADDR_UNICAST))
2002 if (cfg->fc_table) {
2003 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2006 if (grt->rt6i_flags & RTF_GATEWAY ||
2007 (dev && dev != grt->dst.dev)) {
2015 grt = rt6_lookup(net, gw_addr, NULL,
2016 cfg->fc_ifindex, 1);
2018 err = -EHOSTUNREACH;
2022 if (dev != grt->dst.dev) {
2028 idev = grt->rt6i_idev;
2030 in6_dev_hold(grt->rt6i_idev);
2032 if (!(grt->rt6i_flags & RTF_GATEWAY))
2040 if (!dev || (dev->flags & IFF_LOOPBACK))
2048 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2049 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2053 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2054 rt->rt6i_prefsrc.plen = 128;
2056 rt->rt6i_prefsrc.plen = 0;
2058 rt->rt6i_flags = cfg->fc_flags;
2062 rt->rt6i_idev = idev;
2063 rt->rt6i_table = table;
2065 cfg->fc_nlinfo.nl_net = dev_net(dev);
2076 return ERR_PTR(err);
2079 int ip6_route_add(struct fib6_config *cfg)
2081 struct mx6_config mxc = { .mx = NULL, };
2082 struct rt6_info *rt;
2085 rt = ip6_route_info_create(cfg);
2092 err = ip6_convert_metrics(&mxc, cfg);
2096 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2108 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2111 struct fib6_table *table;
2112 struct net *net = dev_net(rt->dst.dev);
2114 if (rt == net->ipv6.ip6_null_entry ||
2115 rt->dst.flags & DST_NOCACHE) {
2120 table = rt->rt6i_table;
2121 write_lock_bh(&table->tb6_lock);
2122 err = fib6_del(rt, info);
2123 write_unlock_bh(&table->tb6_lock);
2130 int ip6_del_rt(struct rt6_info *rt)
2132 struct nl_info info = {
2133 .nl_net = dev_net(rt->dst.dev),
2135 return __ip6_del_rt(rt, &info);
2138 static int ip6_route_del(struct fib6_config *cfg)
2140 struct fib6_table *table;
2141 struct fib6_node *fn;
2142 struct rt6_info *rt;
2145 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2149 read_lock_bh(&table->tb6_lock);
2151 fn = fib6_locate(&table->tb6_root,
2152 &cfg->fc_dst, cfg->fc_dst_len,
2153 &cfg->fc_src, cfg->fc_src_len);
2156 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2157 if ((rt->rt6i_flags & RTF_CACHE) &&
2158 !(cfg->fc_flags & RTF_CACHE))
2160 if (cfg->fc_ifindex &&
2162 rt->dst.dev->ifindex != cfg->fc_ifindex))
2164 if (cfg->fc_flags & RTF_GATEWAY &&
2165 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2167 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2170 read_unlock_bh(&table->tb6_lock);
2172 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2175 read_unlock_bh(&table->tb6_lock);
2180 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2182 struct netevent_redirect netevent;
2183 struct rt6_info *rt, *nrt = NULL;
2184 struct ndisc_options ndopts;
2185 struct inet6_dev *in6_dev;
2186 struct neighbour *neigh;
2188 int optlen, on_link;
2191 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2192 optlen -= sizeof(*msg);
2195 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2199 msg = (struct rd_msg *)icmp6_hdr(skb);
2201 if (ipv6_addr_is_multicast(&msg->dest)) {
2202 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2207 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2209 } else if (ipv6_addr_type(&msg->target) !=
2210 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2211 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2215 in6_dev = __in6_dev_get(skb->dev);
2218 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2222 * The IP source address of the Redirect MUST be the same as the current
2223 * first-hop router for the specified ICMP Destination Address.
2226 if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2227 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2232 if (ndopts.nd_opts_tgt_lladdr) {
2233 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2236 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2241 rt = (struct rt6_info *) dst;
2242 if (rt->rt6i_flags & RTF_REJECT) {
2243 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2247 /* Redirect received -> path was valid.
2248 * Look, redirects are sent only in response to data packets,
2249 * so that this nexthop apparently is reachable. --ANK
2251 dst_confirm(&rt->dst);
2253 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2258 * We have finally decided to accept it.
2261 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2262 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2263 NEIGH_UPDATE_F_OVERRIDE|
2264 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2265 NEIGH_UPDATE_F_ISROUTER)),
2266 NDISC_REDIRECT, &ndopts);
2268 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2272 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2274 nrt->rt6i_flags &= ~RTF_GATEWAY;
2276 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2278 if (ip6_ins_rt(nrt))
2281 netevent.old = &rt->dst;
2282 netevent.new = &nrt->dst;
2283 netevent.daddr = &msg->dest;
2284 netevent.neigh = neigh;
2285 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2287 if (rt->rt6i_flags & RTF_CACHE) {
2288 rt = (struct rt6_info *) dst_clone(&rt->dst);
2293 neigh_release(neigh);
2297 * Misc support functions
2300 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2302 BUG_ON(from->dst.from);
2304 rt->rt6i_flags &= ~RTF_EXPIRES;
2305 dst_hold(&from->dst);
2306 rt->dst.from = &from->dst;
2307 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2310 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2312 rt->dst.input = ort->dst.input;
2313 rt->dst.output = ort->dst.output;
2314 rt->rt6i_dst = ort->rt6i_dst;
2315 rt->dst.error = ort->dst.error;
2316 rt->rt6i_idev = ort->rt6i_idev;
2318 in6_dev_hold(rt->rt6i_idev);
2319 rt->dst.lastuse = jiffies;
2320 rt->rt6i_gateway = ort->rt6i_gateway;
2321 rt->rt6i_flags = ort->rt6i_flags;
2322 rt6_set_from(rt, ort);
2323 rt->rt6i_metric = ort->rt6i_metric;
2324 #ifdef CONFIG_IPV6_SUBTREES
2325 rt->rt6i_src = ort->rt6i_src;
2327 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2328 rt->rt6i_table = ort->rt6i_table;
2329 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2332 #ifdef CONFIG_IPV6_ROUTE_INFO
2333 static struct rt6_info *rt6_get_route_info(struct net *net,
2334 const struct in6_addr *prefix, int prefixlen,
2335 const struct in6_addr *gwaddr,
2336 struct net_device *dev)
2338 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2339 int ifindex = dev->ifindex;
2340 struct fib6_node *fn;
2341 struct rt6_info *rt = NULL;
2342 struct fib6_table *table;
2344 table = fib6_get_table(net, tb_id);
2348 read_lock_bh(&table->tb6_lock);
2349 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2353 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2354 if (rt->dst.dev->ifindex != ifindex)
2356 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2358 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2364 read_unlock_bh(&table->tb6_lock);
2368 static struct rt6_info *rt6_add_route_info(struct net *net,
2369 const struct in6_addr *prefix, int prefixlen,
2370 const struct in6_addr *gwaddr,
2371 struct net_device *dev,
2374 struct fib6_config cfg = {
2375 .fc_metric = IP6_RT_PRIO_USER,
2376 .fc_ifindex = dev->ifindex,
2377 .fc_dst_len = prefixlen,
2378 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2379 RTF_UP | RTF_PREF(pref),
2380 .fc_nlinfo.portid = 0,
2381 .fc_nlinfo.nlh = NULL,
2382 .fc_nlinfo.nl_net = net,
2385 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2386 cfg.fc_dst = *prefix;
2387 cfg.fc_gateway = *gwaddr;
2389 /* We should treat it as a default route if prefix length is 0. */
2391 cfg.fc_flags |= RTF_DEFAULT;
2393 ip6_route_add(&cfg);
2395 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2399 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2401 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2402 struct rt6_info *rt;
2403 struct fib6_table *table;
2405 table = fib6_get_table(dev_net(dev), tb_id);
2409 read_lock_bh(&table->tb6_lock);
2410 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2411 if (dev == rt->dst.dev &&
2412 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2413 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2418 read_unlock_bh(&table->tb6_lock);
2422 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2423 struct net_device *dev,
2426 struct fib6_config cfg = {
2427 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2428 .fc_metric = IP6_RT_PRIO_USER,
2429 .fc_ifindex = dev->ifindex,
2430 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2431 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2432 .fc_nlinfo.portid = 0,
2433 .fc_nlinfo.nlh = NULL,
2434 .fc_nlinfo.nl_net = dev_net(dev),
2437 cfg.fc_gateway = *gwaddr;
2439 if (!ip6_route_add(&cfg)) {
2440 struct fib6_table *table;
2442 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2444 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2447 return rt6_get_dflt_router(gwaddr, dev);
2450 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2452 struct rt6_info *rt;
2455 read_lock_bh(&table->tb6_lock);
2456 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2457 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2458 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2460 read_unlock_bh(&table->tb6_lock);
2465 read_unlock_bh(&table->tb6_lock);
2467 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2470 void rt6_purge_dflt_routers(struct net *net)
2472 struct fib6_table *table;
2473 struct hlist_head *head;
2478 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2479 head = &net->ipv6.fib_table_hash[h];
2480 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2481 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2482 __rt6_purge_dflt_routers(table);
2489 static void rtmsg_to_fib6_config(struct net *net,
2490 struct in6_rtmsg *rtmsg,
2491 struct fib6_config *cfg)
2493 memset(cfg, 0, sizeof(*cfg));
2495 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2497 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2498 cfg->fc_metric = rtmsg->rtmsg_metric;
2499 cfg->fc_expires = rtmsg->rtmsg_info;
2500 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2501 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2502 cfg->fc_flags = rtmsg->rtmsg_flags;
2504 cfg->fc_nlinfo.nl_net = net;
2506 cfg->fc_dst = rtmsg->rtmsg_dst;
2507 cfg->fc_src = rtmsg->rtmsg_src;
2508 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2511 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2513 struct fib6_config cfg;
2514 struct in6_rtmsg rtmsg;
2518 case SIOCADDRT: /* Add a route */
2519 case SIOCDELRT: /* Delete a route */
2520 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2522 err = copy_from_user(&rtmsg, arg,
2523 sizeof(struct in6_rtmsg));
2527 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2532 err = ip6_route_add(&cfg);
2535 err = ip6_route_del(&cfg);
2549 * Drop the packet on the floor
2552 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2555 struct dst_entry *dst = skb_dst(skb);
2556 switch (ipstats_mib_noroutes) {
2557 case IPSTATS_MIB_INNOROUTES:
2558 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2559 if (type == IPV6_ADDR_ANY) {
2560 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2561 IPSTATS_MIB_INADDRERRORS);
2565 case IPSTATS_MIB_OUTNOROUTES:
2566 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2567 ipstats_mib_noroutes);
2570 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2575 static int ip6_pkt_discard(struct sk_buff *skb)
2577 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2580 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2582 skb->dev = skb_dst(skb)->dev;
2583 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2586 static int ip6_pkt_prohibit(struct sk_buff *skb)
2588 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2591 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2593 skb->dev = skb_dst(skb)->dev;
2594 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2598 * Allocate a dst for local (unicast / anycast) address.
2601 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2602 const struct in6_addr *addr,
2606 struct net *net = dev_net(idev->dev);
2607 struct net_device *dev = net->loopback_dev;
2608 struct rt6_info *rt;
2610 /* use L3 Master device as loopback for host routes if device
2611 * is enslaved and address is not link local or multicast
2613 if (!rt6_need_strict(addr))
2614 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2616 rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2618 return ERR_PTR(-ENOMEM);
2622 rt->dst.flags |= DST_HOST;
2623 rt->dst.input = ip6_input;
2624 rt->dst.output = ip6_output;
2625 rt->rt6i_idev = idev;
2627 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2629 rt->rt6i_flags |= RTF_ANYCAST;
2631 rt->rt6i_flags |= RTF_LOCAL;
2633 rt->rt6i_gateway = *addr;
2634 rt->rt6i_dst.addr = *addr;
2635 rt->rt6i_dst.plen = 128;
2636 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2637 rt->rt6i_table = fib6_get_table(net, tb_id);
2638 rt->dst.flags |= DST_NOCACHE;
2640 atomic_set(&rt->dst.__refcnt, 1);
2645 /* remove deleted ip from prefsrc entries */
2646 struct arg_dev_net_ip {
2647 struct net_device *dev;
2649 struct in6_addr *addr;
2652 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2654 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2655 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2656 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2658 if (((void *)rt->dst.dev == dev || !dev) &&
2659 rt != net->ipv6.ip6_null_entry &&
2660 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2661 /* remove prefsrc entry */
2662 rt->rt6i_prefsrc.plen = 0;
2667 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2669 struct net *net = dev_net(ifp->idev->dev);
2670 struct arg_dev_net_ip adni = {
2671 .dev = ifp->idev->dev,
2675 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2678 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2679 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2681 /* Remove routers and update dst entries when gateway turn into host. */
2682 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2684 struct in6_addr *gateway = (struct in6_addr *)arg;
2686 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2687 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2688 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2694 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2696 fib6_clean_all(net, fib6_clean_tohost, gateway);
2699 struct arg_dev_net {
2700 struct net_device *dev;
2704 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2706 const struct arg_dev_net *adn = arg;
2707 const struct net_device *dev = adn->dev;
2709 if ((rt->dst.dev == dev || !dev) &&
2710 rt != adn->net->ipv6.ip6_null_entry)
2716 void rt6_ifdown(struct net *net, struct net_device *dev)
2718 struct arg_dev_net adn = {
2723 fib6_clean_all(net, fib6_ifdown, &adn);
2724 icmp6_clean_all(fib6_ifdown, &adn);
2726 rt6_uncached_list_flush_dev(net, dev);
2729 struct rt6_mtu_change_arg {
2730 struct net_device *dev;
2734 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2736 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2737 struct inet6_dev *idev;
2739 /* In IPv6 pmtu discovery is not optional,
2740 so that RTAX_MTU lock cannot disable it.
2741 We still use this lock to block changes
2742 caused by addrconf/ndisc.
2745 idev = __in6_dev_get(arg->dev);
2749 /* For administrative MTU increase, there is no way to discover
2750 IPv6 PMTU increase, so PMTU increase should be updated here.
2751 Since RFC 1981 doesn't include administrative MTU increase
2752 update PMTU increase is a MUST. (i.e. jumbo frame)
2755 If new MTU is less than route PMTU, this new MTU will be the
2756 lowest MTU in the path, update the route PMTU to reflect PMTU
2757 decreases; if new MTU is greater than route PMTU, and the
2758 old MTU is the lowest MTU in the path, update the route PMTU
2759 to reflect the increase. In this case if the other nodes' MTU
2760 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2763 if (rt->dst.dev == arg->dev &&
2764 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2765 if (rt->rt6i_flags & RTF_CACHE) {
2766 /* For RTF_CACHE with rt6i_pmtu == 0
2767 * (i.e. a redirected route),
2768 * the metrics of its rt->dst.from has already
2771 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2772 rt->rt6i_pmtu = arg->mtu;
2773 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2774 (dst_mtu(&rt->dst) < arg->mtu &&
2775 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2776 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2782 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2784 struct rt6_mtu_change_arg arg = {
2789 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2792 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2793 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2794 [RTA_OIF] = { .type = NLA_U32 },
2795 [RTA_IIF] = { .type = NLA_U32 },
2796 [RTA_PRIORITY] = { .type = NLA_U32 },
2797 [RTA_METRICS] = { .type = NLA_NESTED },
2798 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2799 [RTA_PREF] = { .type = NLA_U8 },
2800 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2801 [RTA_ENCAP] = { .type = NLA_NESTED },
2802 [RTA_EXPIRES] = { .type = NLA_U32 },
2805 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2806 struct fib6_config *cfg)
2809 struct nlattr *tb[RTA_MAX+1];
2813 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2818 rtm = nlmsg_data(nlh);
2819 memset(cfg, 0, sizeof(*cfg));
2821 cfg->fc_table = rtm->rtm_table;
2822 cfg->fc_dst_len = rtm->rtm_dst_len;
2823 cfg->fc_src_len = rtm->rtm_src_len;
2824 cfg->fc_flags = RTF_UP;
2825 cfg->fc_protocol = rtm->rtm_protocol;
2826 cfg->fc_type = rtm->rtm_type;
2828 if (rtm->rtm_type == RTN_UNREACHABLE ||
2829 rtm->rtm_type == RTN_BLACKHOLE ||
2830 rtm->rtm_type == RTN_PROHIBIT ||
2831 rtm->rtm_type == RTN_THROW)
2832 cfg->fc_flags |= RTF_REJECT;
2834 if (rtm->rtm_type == RTN_LOCAL)
2835 cfg->fc_flags |= RTF_LOCAL;
2837 if (rtm->rtm_flags & RTM_F_CLONED)
2838 cfg->fc_flags |= RTF_CACHE;
2840 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2841 cfg->fc_nlinfo.nlh = nlh;
2842 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2844 if (tb[RTA_GATEWAY]) {
2845 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2846 cfg->fc_flags |= RTF_GATEWAY;
2850 int plen = (rtm->rtm_dst_len + 7) >> 3;
2852 if (nla_len(tb[RTA_DST]) < plen)
2855 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2859 int plen = (rtm->rtm_src_len + 7) >> 3;
2861 if (nla_len(tb[RTA_SRC]) < plen)
2864 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2867 if (tb[RTA_PREFSRC])
2868 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2871 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2873 if (tb[RTA_PRIORITY])
2874 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2876 if (tb[RTA_METRICS]) {
2877 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2878 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2882 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2884 if (tb[RTA_MULTIPATH]) {
2885 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2886 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2890 pref = nla_get_u8(tb[RTA_PREF]);
2891 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2892 pref != ICMPV6_ROUTER_PREF_HIGH)
2893 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2894 cfg->fc_flags |= RTF_PREF(pref);
2898 cfg->fc_encap = tb[RTA_ENCAP];
2900 if (tb[RTA_ENCAP_TYPE])
2901 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2903 if (tb[RTA_EXPIRES]) {
2904 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2906 if (addrconf_finite_timeout(timeout)) {
2907 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2908 cfg->fc_flags |= RTF_EXPIRES;
2918 struct rt6_info *rt6_info;
2919 struct fib6_config r_cfg;
2920 struct mx6_config mxc;
2921 struct list_head next;
2924 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2928 list_for_each_entry(nh, rt6_nh_list, next) {
2929 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2930 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2931 nh->r_cfg.fc_ifindex);
2935 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2936 struct rt6_info *rt, struct fib6_config *r_cfg)
2939 struct rt6_info *rtnh;
2942 list_for_each_entry(nh, rt6_nh_list, next) {
2943 /* check if rt6_info already exists */
2944 rtnh = nh->rt6_info;
2946 if (rtnh->dst.dev == rt->dst.dev &&
2947 rtnh->rt6i_idev == rt->rt6i_idev &&
2948 ipv6_addr_equal(&rtnh->rt6i_gateway,
2953 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2957 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2962 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2963 list_add_tail(&nh->next, rt6_nh_list);
2968 static int ip6_route_multipath_add(struct fib6_config *cfg)
2970 struct fib6_config r_cfg;
2971 struct rtnexthop *rtnh;
2972 struct rt6_info *rt;
2973 struct rt6_nh *err_nh;
2974 struct rt6_nh *nh, *nh_safe;
2979 int replace = (cfg->fc_nlinfo.nlh &&
2980 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2981 LIST_HEAD(rt6_nh_list);
2983 remaining = cfg->fc_mp_len;
2984 rtnh = (struct rtnexthop *)cfg->fc_mp;
2986 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2987 * rt6_info structs per nexthop
2989 while (rtnh_ok(rtnh, remaining)) {
2990 memcpy(&r_cfg, cfg, sizeof(*cfg));
2991 if (rtnh->rtnh_ifindex)
2992 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2994 attrlen = rtnh_attrlen(rtnh);
2996 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2998 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3000 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3001 r_cfg.fc_flags |= RTF_GATEWAY;
3003 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3004 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3006 r_cfg.fc_encap_type = nla_get_u16(nla);
3009 rt = ip6_route_info_create(&r_cfg);
3016 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3022 rtnh = rtnh_next(rtnh, &remaining);
3026 list_for_each_entry(nh, &rt6_nh_list, next) {
3027 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
3028 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3029 nh->rt6_info = NULL;
3032 ip6_print_replace_route_err(&rt6_nh_list);
3037 /* Because each route is added like a single route we remove
3038 * these flags after the first nexthop: if there is a collision,
3039 * we have already failed to add the first nexthop:
3040 * fib6_add_rt2node() has rejected it; when replacing, old
3041 * nexthops have been replaced by first new, the rest should
3044 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3052 /* Delete routes that were already added */
3053 list_for_each_entry(nh, &rt6_nh_list, next) {
3056 ip6_route_del(&nh->r_cfg);
3060 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3062 dst_free(&nh->rt6_info->dst);
3064 list_del(&nh->next);
3071 static int ip6_route_multipath_del(struct fib6_config *cfg)
3073 struct fib6_config r_cfg;
3074 struct rtnexthop *rtnh;
3077 int err = 1, last_err = 0;
3079 remaining = cfg->fc_mp_len;
3080 rtnh = (struct rtnexthop *)cfg->fc_mp;
3082 /* Parse a Multipath Entry */
3083 while (rtnh_ok(rtnh, remaining)) {
3084 memcpy(&r_cfg, cfg, sizeof(*cfg));
3085 if (rtnh->rtnh_ifindex)
3086 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3088 attrlen = rtnh_attrlen(rtnh);
3090 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3092 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3094 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3095 r_cfg.fc_flags |= RTF_GATEWAY;
3098 err = ip6_route_del(&r_cfg);
3102 rtnh = rtnh_next(rtnh, &remaining);
3108 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3110 struct fib6_config cfg;
3113 err = rtm_to_fib6_config(skb, nlh, &cfg);
3118 return ip6_route_multipath_del(&cfg);
3120 return ip6_route_del(&cfg);
3123 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3125 struct fib6_config cfg;
3128 err = rtm_to_fib6_config(skb, nlh, &cfg);
3133 return ip6_route_multipath_add(&cfg);
3135 return ip6_route_add(&cfg);
3138 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3140 return NLMSG_ALIGN(sizeof(struct rtmsg))
3141 + nla_total_size(16) /* RTA_SRC */
3142 + nla_total_size(16) /* RTA_DST */
3143 + nla_total_size(16) /* RTA_GATEWAY */
3144 + nla_total_size(16) /* RTA_PREFSRC */
3145 + nla_total_size(4) /* RTA_TABLE */
3146 + nla_total_size(4) /* RTA_IIF */
3147 + nla_total_size(4) /* RTA_OIF */
3148 + nla_total_size(4) /* RTA_PRIORITY */
3149 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3150 + nla_total_size(sizeof(struct rta_cacheinfo))
3151 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3152 + nla_total_size(1) /* RTA_PREF */
3153 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3156 static int rt6_fill_node(struct net *net,
3157 struct sk_buff *skb, struct rt6_info *rt,
3158 struct in6_addr *dst, struct in6_addr *src,
3159 int iif, int type, u32 portid, u32 seq,
3160 int prefix, int nowait, unsigned int flags)
3162 u32 metrics[RTAX_MAX];
3164 struct nlmsghdr *nlh;
3168 if (prefix) { /* user wants prefix routes only */
3169 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3170 /* success since this is not a prefix route */
3175 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3179 rtm = nlmsg_data(nlh);
3180 rtm->rtm_family = AF_INET6;
3181 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3182 rtm->rtm_src_len = rt->rt6i_src.plen;
3185 table = rt->rt6i_table->tb6_id;
3187 table = RT6_TABLE_UNSPEC;
3188 rtm->rtm_table = table;
3189 if (nla_put_u32(skb, RTA_TABLE, table))
3190 goto nla_put_failure;
3191 if (rt->rt6i_flags & RTF_REJECT) {
3192 switch (rt->dst.error) {
3194 rtm->rtm_type = RTN_BLACKHOLE;
3197 rtm->rtm_type = RTN_PROHIBIT;
3200 rtm->rtm_type = RTN_THROW;
3203 rtm->rtm_type = RTN_UNREACHABLE;
3207 else if (rt->rt6i_flags & RTF_LOCAL)
3208 rtm->rtm_type = RTN_LOCAL;
3209 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3210 rtm->rtm_type = RTN_LOCAL;
3212 rtm->rtm_type = RTN_UNICAST;
3214 if (!netif_carrier_ok(rt->dst.dev)) {
3215 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3216 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3217 rtm->rtm_flags |= RTNH_F_DEAD;
3219 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3220 rtm->rtm_protocol = rt->rt6i_protocol;
3221 if (rt->rt6i_flags & RTF_DYNAMIC)
3222 rtm->rtm_protocol = RTPROT_REDIRECT;
3223 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3224 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3225 rtm->rtm_protocol = RTPROT_RA;
3227 rtm->rtm_protocol = RTPROT_KERNEL;
3230 if (rt->rt6i_flags & RTF_CACHE)
3231 rtm->rtm_flags |= RTM_F_CLONED;
3234 if (nla_put_in6_addr(skb, RTA_DST, dst))
3235 goto nla_put_failure;
3236 rtm->rtm_dst_len = 128;
3237 } else if (rtm->rtm_dst_len)
3238 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3239 goto nla_put_failure;
3240 #ifdef CONFIG_IPV6_SUBTREES
3242 if (nla_put_in6_addr(skb, RTA_SRC, src))
3243 goto nla_put_failure;
3244 rtm->rtm_src_len = 128;
3245 } else if (rtm->rtm_src_len &&
3246 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3247 goto nla_put_failure;
3250 #ifdef CONFIG_IPV6_MROUTE
3251 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3252 int err = ip6mr_get_route(net, skb, rtm, nowait,
3259 goto nla_put_failure;
3261 if (err == -EMSGSIZE)
3262 goto nla_put_failure;
3267 if (nla_put_u32(skb, RTA_IIF, iif))
3268 goto nla_put_failure;
3270 struct in6_addr saddr_buf;
3271 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3272 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3273 goto nla_put_failure;
3276 if (rt->rt6i_prefsrc.plen) {
3277 struct in6_addr saddr_buf;
3278 saddr_buf = rt->rt6i_prefsrc.addr;
3279 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3280 goto nla_put_failure;
3283 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3285 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3286 if (rtnetlink_put_metrics(skb, metrics) < 0)
3287 goto nla_put_failure;
3289 if (rt->rt6i_flags & RTF_GATEWAY) {
3290 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3291 goto nla_put_failure;
3295 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3296 goto nla_put_failure;
3297 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3298 goto nla_put_failure;
3300 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3302 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3303 goto nla_put_failure;
3305 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3306 goto nla_put_failure;
3308 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3310 nlmsg_end(skb, nlh);
3314 nlmsg_cancel(skb, nlh);
3318 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3320 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3323 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3324 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3325 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3329 return rt6_fill_node(arg->net,
3330 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3331 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3332 prefix, 0, NLM_F_MULTI);
3335 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3337 struct net *net = sock_net(in_skb->sk);
3338 struct nlattr *tb[RTA_MAX+1];
3339 struct rt6_info *rt;
3340 struct sk_buff *skb;
3343 int err, iif = 0, oif = 0;
3345 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3350 memset(&fl6, 0, sizeof(fl6));
3351 rtm = nlmsg_data(nlh);
3352 fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3355 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3358 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3362 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3365 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3369 iif = nla_get_u32(tb[RTA_IIF]);
3372 oif = nla_get_u32(tb[RTA_OIF]);
3375 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3378 struct net_device *dev;
3381 dev = __dev_get_by_index(net, iif);
3387 fl6.flowi6_iif = iif;
3389 if (!ipv6_addr_any(&fl6.saddr))
3390 flags |= RT6_LOOKUP_F_HAS_SADDR;
3392 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3395 fl6.flowi6_oif = oif;
3397 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3400 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3407 /* Reserve room for dummy headers, this skb can pass
3408 through good chunk of routing engine.
3410 skb_reset_mac_header(skb);
3411 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3413 skb_dst_set(skb, &rt->dst);
3415 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3416 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3417 nlh->nlmsg_seq, 0, 0, 0);
3423 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3428 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3429 unsigned int nlm_flags)
3431 struct sk_buff *skb;
3432 struct net *net = info->nl_net;
3437 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3439 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3443 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3444 event, info->portid, seq, 0, 0, nlm_flags);
3446 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3447 WARN_ON(err == -EMSGSIZE);
3451 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3452 info->nlh, gfp_any());
3456 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3459 static int ip6_route_dev_notify(struct notifier_block *this,
3460 unsigned long event, void *ptr)
3462 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3463 struct net *net = dev_net(dev);
3465 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3466 net->ipv6.ip6_null_entry->dst.dev = dev;
3467 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3468 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3469 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3470 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3471 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3472 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3483 #ifdef CONFIG_PROC_FS
3485 static const struct file_operations ipv6_route_proc_fops = {
3486 .owner = THIS_MODULE,
3487 .open = ipv6_route_open,
3489 .llseek = seq_lseek,
3490 .release = seq_release_net,
3493 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3495 struct net *net = (struct net *)seq->private;
3496 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3497 net->ipv6.rt6_stats->fib_nodes,
3498 net->ipv6.rt6_stats->fib_route_nodes,
3499 net->ipv6.rt6_stats->fib_rt_alloc,
3500 net->ipv6.rt6_stats->fib_rt_entries,
3501 net->ipv6.rt6_stats->fib_rt_cache,
3502 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3503 net->ipv6.rt6_stats->fib_discarded_routes);
3508 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3510 return single_open_net(inode, file, rt6_stats_seq_show);
3513 static const struct file_operations rt6_stats_seq_fops = {
3514 .owner = THIS_MODULE,
3515 .open = rt6_stats_seq_open,
3517 .llseek = seq_lseek,
3518 .release = single_release_net,
3520 #endif /* CONFIG_PROC_FS */
3522 #ifdef CONFIG_SYSCTL
3525 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3526 void __user *buffer, size_t *lenp, loff_t *ppos)
3533 net = (struct net *)ctl->extra1;
3534 delay = net->ipv6.sysctl.flush_delay;
3535 proc_dointvec(ctl, write, buffer, lenp, ppos);
3536 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3540 struct ctl_table ipv6_route_table_template[] = {
3542 .procname = "flush",
3543 .data = &init_net.ipv6.sysctl.flush_delay,
3544 .maxlen = sizeof(int),
3546 .proc_handler = ipv6_sysctl_rtcache_flush
3549 .procname = "gc_thresh",
3550 .data = &ip6_dst_ops_template.gc_thresh,
3551 .maxlen = sizeof(int),
3553 .proc_handler = proc_dointvec,
3556 .procname = "max_size",
3557 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3558 .maxlen = sizeof(int),
3560 .proc_handler = proc_dointvec,
3563 .procname = "gc_min_interval",
3564 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3565 .maxlen = sizeof(int),
3567 .proc_handler = proc_dointvec_jiffies,
3570 .procname = "gc_timeout",
3571 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3572 .maxlen = sizeof(int),
3574 .proc_handler = proc_dointvec_jiffies,
3577 .procname = "gc_interval",
3578 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3579 .maxlen = sizeof(int),
3581 .proc_handler = proc_dointvec_jiffies,
3584 .procname = "gc_elasticity",
3585 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3586 .maxlen = sizeof(int),
3588 .proc_handler = proc_dointvec,
3591 .procname = "mtu_expires",
3592 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3593 .maxlen = sizeof(int),
3595 .proc_handler = proc_dointvec_jiffies,
3598 .procname = "min_adv_mss",
3599 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3600 .maxlen = sizeof(int),
3602 .proc_handler = proc_dointvec,
3605 .procname = "gc_min_interval_ms",
3606 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3607 .maxlen = sizeof(int),
3609 .proc_handler = proc_dointvec_ms_jiffies,
3614 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3616 struct ctl_table *table;
3618 table = kmemdup(ipv6_route_table_template,
3619 sizeof(ipv6_route_table_template),
3623 table[0].data = &net->ipv6.sysctl.flush_delay;
3624 table[0].extra1 = net;
3625 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3626 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3627 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3628 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3629 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3630 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3631 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3632 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3633 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3635 /* Don't export sysctls to unprivileged users */
3636 if (net->user_ns != &init_user_ns)
3637 table[0].procname = NULL;
3644 static int __net_init ip6_route_net_init(struct net *net)
3648 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3649 sizeof(net->ipv6.ip6_dst_ops));
3651 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3652 goto out_ip6_dst_ops;
3654 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3655 sizeof(*net->ipv6.ip6_null_entry),
3657 if (!net->ipv6.ip6_null_entry)
3658 goto out_ip6_dst_entries;
3659 net->ipv6.ip6_null_entry->dst.path =
3660 (struct dst_entry *)net->ipv6.ip6_null_entry;
3661 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3662 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3663 ip6_template_metrics, true);
3665 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3666 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3667 sizeof(*net->ipv6.ip6_prohibit_entry),
3669 if (!net->ipv6.ip6_prohibit_entry)
3670 goto out_ip6_null_entry;
3671 net->ipv6.ip6_prohibit_entry->dst.path =
3672 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3673 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3674 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3675 ip6_template_metrics, true);
3677 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3678 sizeof(*net->ipv6.ip6_blk_hole_entry),
3680 if (!net->ipv6.ip6_blk_hole_entry)
3681 goto out_ip6_prohibit_entry;
3682 net->ipv6.ip6_blk_hole_entry->dst.path =
3683 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3684 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3685 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3686 ip6_template_metrics, true);
3689 net->ipv6.sysctl.flush_delay = 0;
3690 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3691 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3692 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3693 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3694 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3695 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3696 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3698 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3704 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3705 out_ip6_prohibit_entry:
3706 kfree(net->ipv6.ip6_prohibit_entry);
3708 kfree(net->ipv6.ip6_null_entry);
3710 out_ip6_dst_entries:
3711 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3716 static void __net_exit ip6_route_net_exit(struct net *net)
3718 kfree(net->ipv6.ip6_null_entry);
3719 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3720 kfree(net->ipv6.ip6_prohibit_entry);
3721 kfree(net->ipv6.ip6_blk_hole_entry);
3723 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3726 static int __net_init ip6_route_net_init_late(struct net *net)
3728 #ifdef CONFIG_PROC_FS
3729 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3730 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3735 static void __net_exit ip6_route_net_exit_late(struct net *net)
3737 #ifdef CONFIG_PROC_FS
3738 remove_proc_entry("ipv6_route", net->proc_net);
3739 remove_proc_entry("rt6_stats", net->proc_net);
3743 static struct pernet_operations ip6_route_net_ops = {
3744 .init = ip6_route_net_init,
3745 .exit = ip6_route_net_exit,
3748 static int __net_init ipv6_inetpeer_init(struct net *net)
3750 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3754 inet_peer_base_init(bp);
3755 net->ipv6.peers = bp;
3759 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3761 struct inet_peer_base *bp = net->ipv6.peers;
3763 net->ipv6.peers = NULL;
3764 inetpeer_invalidate_tree(bp);
3768 static struct pernet_operations ipv6_inetpeer_ops = {
3769 .init = ipv6_inetpeer_init,
3770 .exit = ipv6_inetpeer_exit,
3773 static struct pernet_operations ip6_route_net_late_ops = {
3774 .init = ip6_route_net_init_late,
3775 .exit = ip6_route_net_exit_late,
3778 static struct notifier_block ip6_route_dev_notifier = {
3779 .notifier_call = ip6_route_dev_notify,
3783 int __init ip6_route_init(void)
3789 ip6_dst_ops_template.kmem_cachep =
3790 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3791 SLAB_HWCACHE_ALIGN, NULL);
3792 if (!ip6_dst_ops_template.kmem_cachep)
3795 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3797 goto out_kmem_cache;
3799 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3801 goto out_dst_entries;
3803 ret = register_pernet_subsys(&ip6_route_net_ops);
3805 goto out_register_inetpeer;
3807 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3809 /* Registering of the loopback is done before this portion of code,
3810 * the loopback reference in rt6_info will not be taken, do it
3811 * manually for init_net */
3812 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3813 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3814 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3815 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3816 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3817 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3818 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3822 goto out_register_subsys;
3828 ret = fib6_rules_init();
3832 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3834 goto fib6_rules_init;
3837 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3838 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3839 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3840 goto out_register_late_subsys;
3842 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3844 goto out_register_late_subsys;
3846 for_each_possible_cpu(cpu) {
3847 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3849 INIT_LIST_HEAD(&ul->head);
3850 spin_lock_init(&ul->lock);
3856 out_register_late_subsys:
3857 unregister_pernet_subsys(&ip6_route_net_late_ops);
3859 fib6_rules_cleanup();
3864 out_register_subsys:
3865 unregister_pernet_subsys(&ip6_route_net_ops);
3866 out_register_inetpeer:
3867 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3869 dst_entries_destroy(&ip6_dst_blackhole_ops);
3871 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3875 void ip6_route_cleanup(void)
3877 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3878 unregister_pernet_subsys(&ip6_route_net_late_ops);
3879 fib6_rules_cleanup();
3882 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3883 unregister_pernet_subsys(&ip6_route_net_ops);
3884 dst_entries_destroy(&ip6_dst_blackhole_ops);
3885 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);