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, int ifindex,
107 static struct rt6_info *rt6_get_route_info(struct net *net,
108 const struct in6_addr *prefix, int prefixlen,
109 const struct in6_addr *gwaddr, int ifindex);
112 struct uncached_list {
114 struct list_head head;
117 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
119 static void rt6_uncached_list_add(struct rt6_info *rt)
121 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
123 rt->dst.flags |= DST_NOCACHE;
124 rt->rt6i_uncached_list = ul;
126 spin_lock_bh(&ul->lock);
127 list_add_tail(&rt->rt6i_uncached, &ul->head);
128 spin_unlock_bh(&ul->lock);
131 static void rt6_uncached_list_del(struct rt6_info *rt)
133 if (!list_empty(&rt->rt6i_uncached)) {
134 struct uncached_list *ul = rt->rt6i_uncached_list;
136 spin_lock_bh(&ul->lock);
137 list_del(&rt->rt6i_uncached);
138 spin_unlock_bh(&ul->lock);
142 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
144 struct net_device *loopback_dev = net->loopback_dev;
147 if (dev == loopback_dev)
150 for_each_possible_cpu(cpu) {
151 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
154 spin_lock_bh(&ul->lock);
155 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
156 struct inet6_dev *rt_idev = rt->rt6i_idev;
157 struct net_device *rt_dev = rt->dst.dev;
159 if (rt_idev->dev == dev) {
160 rt->rt6i_idev = in6_dev_get(loopback_dev);
161 in6_dev_put(rt_idev);
165 rt->dst.dev = loopback_dev;
166 dev_hold(rt->dst.dev);
170 spin_unlock_bh(&ul->lock);
174 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
176 return dst_metrics_write_ptr(rt->dst.from);
179 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
181 struct rt6_info *rt = (struct rt6_info *)dst;
183 if (rt->rt6i_flags & RTF_PCPU)
184 return rt6_pcpu_cow_metrics(rt);
185 else if (rt->rt6i_flags & RTF_CACHE)
188 return dst_cow_metrics_generic(dst, old);
191 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
195 struct in6_addr *p = &rt->rt6i_gateway;
197 if (!ipv6_addr_any(p))
198 return (const void *) p;
200 return &ipv6_hdr(skb)->daddr;
204 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
208 struct rt6_info *rt = (struct rt6_info *) dst;
211 daddr = choose_neigh_daddr(rt, skb, daddr);
212 n = __ipv6_neigh_lookup(dst->dev, daddr);
215 return neigh_create(&nd_tbl, daddr, dst->dev);
218 static struct dst_ops ip6_dst_ops_template = {
222 .check = ip6_dst_check,
223 .default_advmss = ip6_default_advmss,
225 .cow_metrics = ipv6_cow_metrics,
226 .destroy = ip6_dst_destroy,
227 .ifdown = ip6_dst_ifdown,
228 .negative_advice = ip6_negative_advice,
229 .link_failure = ip6_link_failure,
230 .update_pmtu = ip6_rt_update_pmtu,
231 .redirect = rt6_do_redirect,
232 .local_out = __ip6_local_out,
233 .neigh_lookup = ip6_neigh_lookup,
236 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
238 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
240 return mtu ? : dst->dev->mtu;
243 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
244 struct sk_buff *skb, u32 mtu)
248 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
253 static struct dst_ops ip6_dst_blackhole_ops = {
255 .destroy = ip6_dst_destroy,
256 .check = ip6_dst_check,
257 .mtu = ip6_blackhole_mtu,
258 .default_advmss = ip6_default_advmss,
259 .update_pmtu = ip6_rt_blackhole_update_pmtu,
260 .redirect = ip6_rt_blackhole_redirect,
261 .cow_metrics = dst_cow_metrics_generic,
262 .neigh_lookup = ip6_neigh_lookup,
265 static const u32 ip6_template_metrics[RTAX_MAX] = {
266 [RTAX_HOPLIMIT - 1] = 0,
269 static const struct rt6_info ip6_null_entry_template = {
271 .__refcnt = ATOMIC_INIT(1),
273 .obsolete = DST_OBSOLETE_FORCE_CHK,
274 .error = -ENETUNREACH,
275 .input = ip6_pkt_discard,
276 .output = ip6_pkt_discard_out,
278 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
279 .rt6i_protocol = RTPROT_KERNEL,
280 .rt6i_metric = ~(u32) 0,
281 .rt6i_ref = ATOMIC_INIT(1),
284 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
286 static const struct rt6_info ip6_prohibit_entry_template = {
288 .__refcnt = ATOMIC_INIT(1),
290 .obsolete = DST_OBSOLETE_FORCE_CHK,
292 .input = ip6_pkt_prohibit,
293 .output = ip6_pkt_prohibit_out,
295 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
296 .rt6i_protocol = RTPROT_KERNEL,
297 .rt6i_metric = ~(u32) 0,
298 .rt6i_ref = ATOMIC_INIT(1),
301 static const struct rt6_info ip6_blk_hole_entry_template = {
303 .__refcnt = ATOMIC_INIT(1),
305 .obsolete = DST_OBSOLETE_FORCE_CHK,
307 .input = dst_discard,
308 .output = dst_discard_out,
310 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
311 .rt6i_protocol = RTPROT_KERNEL,
312 .rt6i_metric = ~(u32) 0,
313 .rt6i_ref = ATOMIC_INIT(1),
318 static void rt6_info_init(struct rt6_info *rt)
320 struct dst_entry *dst = &rt->dst;
322 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
323 INIT_LIST_HEAD(&rt->rt6i_siblings);
324 INIT_LIST_HEAD(&rt->rt6i_uncached);
327 /* allocate dst with ip6_dst_ops */
328 static struct rt6_info *__ip6_dst_alloc(struct net *net,
329 struct net_device *dev,
332 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
333 0, DST_OBSOLETE_FORCE_CHK, flags);
341 struct rt6_info *ip6_dst_alloc(struct net *net,
342 struct net_device *dev,
345 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
348 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
352 for_each_possible_cpu(cpu) {
355 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
356 /* no one shares rt */
360 dst_destroy((struct dst_entry *)rt);
367 EXPORT_SYMBOL(ip6_dst_alloc);
369 static void ip6_dst_destroy(struct dst_entry *dst)
371 struct rt6_info *rt = (struct rt6_info *)dst;
372 struct dst_entry *from = dst->from;
373 struct inet6_dev *idev;
375 dst_destroy_metrics_generic(dst);
376 free_percpu(rt->rt6i_pcpu);
377 rt6_uncached_list_del(rt);
379 idev = rt->rt6i_idev;
381 rt->rt6i_idev = NULL;
389 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
392 struct rt6_info *rt = (struct rt6_info *)dst;
393 struct inet6_dev *idev = rt->rt6i_idev;
394 struct net_device *loopback_dev =
395 dev_net(dev)->loopback_dev;
397 if (dev != loopback_dev) {
398 if (idev && idev->dev == dev) {
399 struct inet6_dev *loopback_idev =
400 in6_dev_get(loopback_dev);
402 rt->rt6i_idev = loopback_idev;
409 static bool __rt6_check_expired(const struct rt6_info *rt)
411 if (rt->rt6i_flags & RTF_EXPIRES)
412 return time_after(jiffies, rt->dst.expires);
417 static bool rt6_check_expired(const struct rt6_info *rt)
419 if (rt->rt6i_flags & RTF_EXPIRES) {
420 if (time_after(jiffies, rt->dst.expires))
422 } else if (rt->dst.from) {
423 return rt6_check_expired((struct rt6_info *) rt->dst.from);
428 /* Multipath route selection:
429 * Hash based function using packet header and flowlabel.
430 * Adapted from fib_info_hashfn()
432 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
433 const struct flowi6 *fl6)
435 return get_hash_from_flowi6(fl6) % candidate_count;
438 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
439 struct flowi6 *fl6, int oif,
442 struct rt6_info *sibling, *next_sibling;
445 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
446 /* Don't change the route, if route_choosen == 0
447 * (siblings does not include ourself)
450 list_for_each_entry_safe(sibling, next_sibling,
451 &match->rt6i_siblings, rt6i_siblings) {
453 if (route_choosen == 0) {
454 if (rt6_score_route(sibling, oif, strict) < 0)
464 * Route lookup. Any table->tb6_lock is implied.
467 static inline struct rt6_info *rt6_device_match(struct net *net,
469 const struct in6_addr *saddr,
473 struct rt6_info *local = NULL;
474 struct rt6_info *sprt;
476 if (!oif && ipv6_addr_any(saddr))
479 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
480 struct net_device *dev = sprt->dst.dev;
483 if (dev->ifindex == oif)
485 if (dev->flags & IFF_LOOPBACK) {
486 if (!sprt->rt6i_idev ||
487 sprt->rt6i_idev->dev->ifindex != oif) {
488 if (flags & RT6_LOOKUP_F_IFACE)
491 local->rt6i_idev->dev->ifindex == oif)
497 if (ipv6_chk_addr(net, saddr, dev,
498 flags & RT6_LOOKUP_F_IFACE))
507 if (flags & RT6_LOOKUP_F_IFACE)
508 return net->ipv6.ip6_null_entry;
514 #ifdef CONFIG_IPV6_ROUTER_PREF
515 struct __rt6_probe_work {
516 struct work_struct work;
517 struct in6_addr target;
518 struct net_device *dev;
521 static void rt6_probe_deferred(struct work_struct *w)
523 struct in6_addr mcaddr;
524 struct __rt6_probe_work *work =
525 container_of(w, struct __rt6_probe_work, work);
527 addrconf_addr_solict_mult(&work->target, &mcaddr);
528 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
533 static void rt6_probe(struct rt6_info *rt)
535 struct __rt6_probe_work *work;
536 struct neighbour *neigh;
538 * Okay, this does not seem to be appropriate
539 * for now, however, we need to check if it
540 * is really so; aka Router Reachability Probing.
542 * Router Reachability Probe MUST be rate-limited
543 * to no more than one per minute.
545 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
548 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
550 if (neigh->nud_state & NUD_VALID)
554 write_lock(&neigh->lock);
555 if (!(neigh->nud_state & NUD_VALID) &&
558 rt->rt6i_idev->cnf.rtr_probe_interval)) {
559 work = kmalloc(sizeof(*work), GFP_ATOMIC);
561 __neigh_set_probe_once(neigh);
563 write_unlock(&neigh->lock);
565 work = kmalloc(sizeof(*work), GFP_ATOMIC);
569 INIT_WORK(&work->work, rt6_probe_deferred);
570 work->target = rt->rt6i_gateway;
571 dev_hold(rt->dst.dev);
572 work->dev = rt->dst.dev;
573 schedule_work(&work->work);
577 rcu_read_unlock_bh();
580 static inline void rt6_probe(struct rt6_info *rt)
586 * Default Router Selection (RFC 2461 6.3.6)
588 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
590 struct net_device *dev = rt->dst.dev;
591 if (!oif || dev->ifindex == oif)
593 if ((dev->flags & IFF_LOOPBACK) &&
594 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
599 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
601 struct neighbour *neigh;
602 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
604 if (rt->rt6i_flags & RTF_NONEXTHOP ||
605 !(rt->rt6i_flags & RTF_GATEWAY))
606 return RT6_NUD_SUCCEED;
609 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
611 read_lock(&neigh->lock);
612 if (neigh->nud_state & NUD_VALID)
613 ret = RT6_NUD_SUCCEED;
614 #ifdef CONFIG_IPV6_ROUTER_PREF
615 else if (!(neigh->nud_state & NUD_FAILED))
616 ret = RT6_NUD_SUCCEED;
618 ret = RT6_NUD_FAIL_PROBE;
620 read_unlock(&neigh->lock);
622 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
623 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
625 rcu_read_unlock_bh();
630 static int rt6_score_route(struct rt6_info *rt, int oif,
635 m = rt6_check_dev(rt, oif);
636 if (!m && (strict & RT6_LOOKUP_F_IFACE))
637 return RT6_NUD_FAIL_HARD;
638 #ifdef CONFIG_IPV6_ROUTER_PREF
639 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
641 if (strict & RT6_LOOKUP_F_REACHABLE) {
642 int n = rt6_check_neigh(rt);
649 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
650 int *mpri, struct rt6_info *match,
654 bool match_do_rr = false;
655 struct inet6_dev *idev = rt->rt6i_idev;
656 struct net_device *dev = rt->dst.dev;
658 if (dev && !netif_carrier_ok(dev) &&
659 idev->cnf.ignore_routes_with_linkdown)
662 if (rt6_check_expired(rt))
665 m = rt6_score_route(rt, oif, strict);
666 if (m == RT6_NUD_FAIL_DO_RR) {
668 m = 0; /* lowest valid score */
669 } else if (m == RT6_NUD_FAIL_HARD) {
673 if (strict & RT6_LOOKUP_F_REACHABLE)
676 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
678 *do_rr = match_do_rr;
686 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
687 struct rt6_info *rr_head,
688 u32 metric, int oif, int strict,
691 struct rt6_info *rt, *match, *cont;
696 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
697 if (rt->rt6i_metric != metric) {
702 match = find_match(rt, oif, strict, &mpri, match, do_rr);
705 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
706 if (rt->rt6i_metric != metric) {
711 match = find_match(rt, oif, strict, &mpri, match, do_rr);
717 for (rt = cont; rt; rt = rt->dst.rt6_next)
718 match = find_match(rt, oif, strict, &mpri, match, do_rr);
723 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
725 struct rt6_info *match, *rt0;
731 fn->rr_ptr = rt0 = fn->leaf;
733 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
737 struct rt6_info *next = rt0->dst.rt6_next;
739 /* no entries matched; do round-robin */
740 if (!next || next->rt6i_metric != rt0->rt6i_metric)
747 net = dev_net(rt0->dst.dev);
748 return match ? match : net->ipv6.ip6_null_entry;
751 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
753 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
756 #ifdef CONFIG_IPV6_ROUTE_INFO
757 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
758 const struct in6_addr *gwaddr)
760 struct net *net = dev_net(dev);
761 struct route_info *rinfo = (struct route_info *) opt;
762 struct in6_addr prefix_buf, *prefix;
764 unsigned long lifetime;
767 if (len < sizeof(struct route_info)) {
771 /* Sanity check for prefix_len and length */
772 if (rinfo->length > 3) {
774 } else if (rinfo->prefix_len > 128) {
776 } else if (rinfo->prefix_len > 64) {
777 if (rinfo->length < 2) {
780 } else if (rinfo->prefix_len > 0) {
781 if (rinfo->length < 1) {
786 pref = rinfo->route_pref;
787 if (pref == ICMPV6_ROUTER_PREF_INVALID)
790 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
792 if (rinfo->length == 3)
793 prefix = (struct in6_addr *)rinfo->prefix;
795 /* this function is safe */
796 ipv6_addr_prefix(&prefix_buf,
797 (struct in6_addr *)rinfo->prefix,
799 prefix = &prefix_buf;
802 if (rinfo->prefix_len == 0)
803 rt = rt6_get_dflt_router(gwaddr, dev);
805 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
806 gwaddr, dev->ifindex);
808 if (rt && !lifetime) {
814 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
817 rt->rt6i_flags = RTF_ROUTEINFO |
818 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
821 if (!addrconf_finite_timeout(lifetime))
822 rt6_clean_expires(rt);
824 rt6_set_expires(rt, jiffies + HZ * lifetime);
832 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
833 struct in6_addr *saddr)
835 struct fib6_node *pn;
837 if (fn->fn_flags & RTN_TL_ROOT)
840 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
841 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
844 if (fn->fn_flags & RTN_RTINFO)
849 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
850 struct fib6_table *table,
851 struct flowi6 *fl6, int flags)
853 struct fib6_node *fn;
856 read_lock_bh(&table->tb6_lock);
857 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
860 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
861 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
862 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
863 if (rt == net->ipv6.ip6_null_entry) {
864 fn = fib6_backtrack(fn, &fl6->saddr);
868 dst_use(&rt->dst, jiffies);
869 read_unlock_bh(&table->tb6_lock);
871 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
877 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
880 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
882 EXPORT_SYMBOL_GPL(ip6_route_lookup);
884 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
885 const struct in6_addr *saddr, int oif, int strict)
887 struct flowi6 fl6 = {
891 struct dst_entry *dst;
892 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
895 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
896 flags |= RT6_LOOKUP_F_HAS_SADDR;
899 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
901 return (struct rt6_info *) dst;
907 EXPORT_SYMBOL(rt6_lookup);
909 /* ip6_ins_rt is called with FREE table->tb6_lock.
910 It takes new route entry, the addition fails by any reason the
911 route is freed. In any case, if caller does not hold it, it may
915 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
916 struct mx6_config *mxc)
919 struct fib6_table *table;
921 table = rt->rt6i_table;
922 write_lock_bh(&table->tb6_lock);
923 err = fib6_add(&table->tb6_root, rt, info, mxc);
924 write_unlock_bh(&table->tb6_lock);
929 int ip6_ins_rt(struct rt6_info *rt)
931 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
932 struct mx6_config mxc = { .mx = NULL, };
934 return __ip6_ins_rt(rt, &info, &mxc);
937 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
938 const struct in6_addr *daddr,
939 const struct in6_addr *saddr)
947 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
948 ort = (struct rt6_info *)ort->dst.from;
950 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
955 ip6_rt_copy_init(rt, ort);
956 rt->rt6i_flags |= RTF_CACHE;
958 rt->dst.flags |= DST_HOST;
959 rt->rt6i_dst.addr = *daddr;
960 rt->rt6i_dst.plen = 128;
962 if (!rt6_is_gw_or_nonexthop(ort)) {
963 if (ort->rt6i_dst.plen != 128 &&
964 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
965 rt->rt6i_flags |= RTF_ANYCAST;
966 #ifdef CONFIG_IPV6_SUBTREES
967 if (rt->rt6i_src.plen && saddr) {
968 rt->rt6i_src.addr = *saddr;
969 rt->rt6i_src.plen = 128;
977 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
979 struct rt6_info *pcpu_rt;
981 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
982 rt->dst.dev, rt->dst.flags);
986 ip6_rt_copy_init(pcpu_rt, rt);
987 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
988 pcpu_rt->rt6i_flags |= RTF_PCPU;
992 /* It should be called with read_lock_bh(&tb6_lock) acquired */
993 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
995 struct rt6_info *pcpu_rt, **p;
997 p = this_cpu_ptr(rt->rt6i_pcpu);
1001 dst_hold(&pcpu_rt->dst);
1002 rt6_dst_from_metrics_check(pcpu_rt);
1007 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1009 struct fib6_table *table = rt->rt6i_table;
1010 struct rt6_info *pcpu_rt, *prev, **p;
1012 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1014 struct net *net = dev_net(rt->dst.dev);
1016 dst_hold(&net->ipv6.ip6_null_entry->dst);
1017 return net->ipv6.ip6_null_entry;
1020 read_lock_bh(&table->tb6_lock);
1021 if (rt->rt6i_pcpu) {
1022 p = this_cpu_ptr(rt->rt6i_pcpu);
1023 prev = cmpxchg(p, NULL, pcpu_rt);
1025 /* If someone did it before us, return prev instead */
1026 dst_destroy(&pcpu_rt->dst);
1030 /* rt has been removed from the fib6 tree
1031 * before we have a chance to acquire the read_lock.
1032 * In this case, don't brother to create a pcpu rt
1033 * since rt is going away anyway. The next
1034 * dst_check() will trigger a re-lookup.
1036 dst_destroy(&pcpu_rt->dst);
1039 dst_hold(&pcpu_rt->dst);
1040 rt6_dst_from_metrics_check(pcpu_rt);
1041 read_unlock_bh(&table->tb6_lock);
1045 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1046 struct flowi6 *fl6, int flags)
1048 struct fib6_node *fn, *saved_fn;
1049 struct rt6_info *rt;
1052 strict |= flags & RT6_LOOKUP_F_IFACE;
1053 if (net->ipv6.devconf_all->forwarding == 0)
1054 strict |= RT6_LOOKUP_F_REACHABLE;
1056 read_lock_bh(&table->tb6_lock);
1058 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1061 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1065 rt = rt6_select(fn, oif, strict);
1066 if (rt->rt6i_nsiblings)
1067 rt = rt6_multipath_select(rt, fl6, oif, strict);
1068 if (rt == net->ipv6.ip6_null_entry) {
1069 fn = fib6_backtrack(fn, &fl6->saddr);
1071 goto redo_rt6_select;
1072 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1073 /* also consider unreachable route */
1074 strict &= ~RT6_LOOKUP_F_REACHABLE;
1076 goto redo_rt6_select;
1081 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1082 dst_use(&rt->dst, jiffies);
1083 read_unlock_bh(&table->tb6_lock);
1085 rt6_dst_from_metrics_check(rt);
1087 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1089 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1090 !(rt->rt6i_flags & RTF_GATEWAY))) {
1091 /* Create a RTF_CACHE clone which will not be
1092 * owned by the fib6 tree. It is for the special case where
1093 * the daddr in the skb during the neighbor look-up is different
1094 * from the fl6->daddr used to look-up route here.
1097 struct rt6_info *uncached_rt;
1099 dst_use(&rt->dst, jiffies);
1100 read_unlock_bh(&table->tb6_lock);
1102 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1103 dst_release(&rt->dst);
1106 rt6_uncached_list_add(uncached_rt);
1108 uncached_rt = net->ipv6.ip6_null_entry;
1110 dst_hold(&uncached_rt->dst);
1112 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1116 /* Get a percpu copy */
1118 struct rt6_info *pcpu_rt;
1120 rt->dst.lastuse = jiffies;
1122 pcpu_rt = rt6_get_pcpu_route(rt);
1125 read_unlock_bh(&table->tb6_lock);
1127 /* We have to do the read_unlock first
1128 * because rt6_make_pcpu_route() may trigger
1129 * ip6_dst_gc() which will take the write_lock.
1132 read_unlock_bh(&table->tb6_lock);
1133 pcpu_rt = rt6_make_pcpu_route(rt);
1134 dst_release(&rt->dst);
1137 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1143 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1144 struct flowi6 *fl6, int flags)
1146 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1149 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1150 struct net_device *dev,
1151 struct flowi6 *fl6, int flags)
1153 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1154 flags |= RT6_LOOKUP_F_IFACE;
1156 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1159 void ip6_route_input(struct sk_buff *skb)
1161 const struct ipv6hdr *iph = ipv6_hdr(skb);
1162 struct net *net = dev_net(skb->dev);
1163 int flags = RT6_LOOKUP_F_HAS_SADDR;
1164 struct ip_tunnel_info *tun_info;
1165 struct flowi6 fl6 = {
1166 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1167 .daddr = iph->daddr,
1168 .saddr = iph->saddr,
1169 .flowlabel = ip6_flowinfo(iph),
1170 .flowi6_mark = skb->mark,
1171 .flowi6_proto = iph->nexthdr,
1174 tun_info = skb_tunnel_info(skb);
1175 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1176 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1178 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1181 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1182 struct flowi6 *fl6, int flags)
1184 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1187 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1188 struct flowi6 *fl6, int flags)
1190 struct dst_entry *dst;
1193 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1197 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1199 any_src = ipv6_addr_any(&fl6->saddr);
1200 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1201 (fl6->flowi6_oif && any_src))
1202 flags |= RT6_LOOKUP_F_IFACE;
1205 flags |= RT6_LOOKUP_F_HAS_SADDR;
1207 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1209 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1211 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1213 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1215 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1216 struct dst_entry *new = NULL;
1218 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1224 new->input = dst_discard;
1225 new->output = dst_discard_out;
1227 dst_copy_metrics(new, &ort->dst);
1228 rt->rt6i_idev = ort->rt6i_idev;
1230 in6_dev_hold(rt->rt6i_idev);
1232 rt->rt6i_gateway = ort->rt6i_gateway;
1233 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1234 rt->rt6i_metric = 0;
1236 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1237 #ifdef CONFIG_IPV6_SUBTREES
1238 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1244 dst_release(dst_orig);
1245 return new ? new : ERR_PTR(-ENOMEM);
1249 * Destination cache support functions
1252 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1255 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1256 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1259 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1261 if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1264 if (rt6_check_expired(rt))
1270 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1272 if (!__rt6_check_expired(rt) &&
1273 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1274 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1280 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1282 struct rt6_info *rt;
1284 rt = (struct rt6_info *) dst;
1286 /* All IPV6 dsts are created with ->obsolete set to the value
1287 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1288 * into this function always.
1291 rt6_dst_from_metrics_check(rt);
1293 if (rt->rt6i_flags & RTF_PCPU ||
1294 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1295 return rt6_dst_from_check(rt, cookie);
1297 return rt6_check(rt, cookie);
1300 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1302 struct rt6_info *rt = (struct rt6_info *) dst;
1305 if (rt->rt6i_flags & RTF_CACHE) {
1306 if (rt6_check_expired(rt)) {
1318 static void ip6_link_failure(struct sk_buff *skb)
1320 struct rt6_info *rt;
1322 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1324 rt = (struct rt6_info *) skb_dst(skb);
1326 if (rt->rt6i_flags & RTF_CACHE) {
1329 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1330 rt->rt6i_node->fn_sernum = -1;
1335 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1337 struct net *net = dev_net(rt->dst.dev);
1339 rt->rt6i_flags |= RTF_MODIFIED;
1340 rt->rt6i_pmtu = mtu;
1341 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1344 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1346 return !(rt->rt6i_flags & RTF_CACHE) &&
1347 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1350 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1351 const struct ipv6hdr *iph, u32 mtu)
1353 struct rt6_info *rt6 = (struct rt6_info *)dst;
1355 if (rt6->rt6i_flags & RTF_LOCAL)
1359 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1360 if (mtu >= dst_mtu(dst))
1363 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1364 rt6_do_update_pmtu(rt6, mtu);
1366 const struct in6_addr *daddr, *saddr;
1367 struct rt6_info *nrt6;
1370 daddr = &iph->daddr;
1371 saddr = &iph->saddr;
1373 daddr = &sk->sk_v6_daddr;
1374 saddr = &inet6_sk(sk)->saddr;
1378 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1380 rt6_do_update_pmtu(nrt6, mtu);
1382 /* ip6_ins_rt(nrt6) will bump the
1383 * rt6->rt6i_node->fn_sernum
1384 * which will fail the next rt6_check() and
1385 * invalidate the sk->sk_dst_cache.
1392 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1393 struct sk_buff *skb, u32 mtu)
1395 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1398 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1401 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1402 struct dst_entry *dst;
1405 memset(&fl6, 0, sizeof(fl6));
1406 fl6.flowi6_oif = oif;
1407 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1408 fl6.daddr = iph->daddr;
1409 fl6.saddr = iph->saddr;
1410 fl6.flowlabel = ip6_flowinfo(iph);
1412 dst = ip6_route_output(net, NULL, &fl6);
1414 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1417 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1419 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1421 struct dst_entry *dst;
1423 ip6_update_pmtu(skb, sock_net(sk), mtu,
1424 sk->sk_bound_dev_if, sk->sk_mark);
1426 dst = __sk_dst_get(sk);
1427 if (!dst || !dst->obsolete ||
1428 dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1432 if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1433 ip6_datagram_dst_update(sk, false);
1436 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1438 /* Handle redirects */
1439 struct ip6rd_flowi {
1441 struct in6_addr gateway;
1444 static struct rt6_info *__ip6_route_redirect(struct net *net,
1445 struct fib6_table *table,
1449 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1450 struct rt6_info *rt;
1451 struct fib6_node *fn;
1453 /* Get the "current" route for this destination and
1454 * check if the redirect has come from approriate router.
1456 * RFC 4861 specifies that redirects should only be
1457 * accepted if they come from the nexthop to the target.
1458 * Due to the way the routes are chosen, this notion
1459 * is a bit fuzzy and one might need to check all possible
1463 read_lock_bh(&table->tb6_lock);
1464 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1466 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1467 if (rt6_check_expired(rt))
1471 if (!(rt->rt6i_flags & RTF_GATEWAY))
1473 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1475 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1481 rt = net->ipv6.ip6_null_entry;
1482 else if (rt->dst.error) {
1483 rt = net->ipv6.ip6_null_entry;
1487 if (rt == net->ipv6.ip6_null_entry) {
1488 fn = fib6_backtrack(fn, &fl6->saddr);
1496 read_unlock_bh(&table->tb6_lock);
1498 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1502 static struct dst_entry *ip6_route_redirect(struct net *net,
1503 const struct flowi6 *fl6,
1504 const struct in6_addr *gateway)
1506 int flags = RT6_LOOKUP_F_HAS_SADDR;
1507 struct ip6rd_flowi rdfl;
1510 rdfl.gateway = *gateway;
1512 return fib6_rule_lookup(net, &rdfl.fl6,
1513 flags, __ip6_route_redirect);
1516 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1518 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1519 struct dst_entry *dst;
1522 memset(&fl6, 0, sizeof(fl6));
1523 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1524 fl6.flowi6_oif = oif;
1525 fl6.flowi6_mark = mark;
1526 fl6.daddr = iph->daddr;
1527 fl6.saddr = iph->saddr;
1528 fl6.flowlabel = ip6_flowinfo(iph);
1530 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1531 rt6_do_redirect(dst, NULL, skb);
1534 EXPORT_SYMBOL_GPL(ip6_redirect);
1536 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1539 const struct ipv6hdr *iph = ipv6_hdr(skb);
1540 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1541 struct dst_entry *dst;
1544 memset(&fl6, 0, sizeof(fl6));
1545 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1546 fl6.flowi6_oif = oif;
1547 fl6.flowi6_mark = mark;
1548 fl6.daddr = msg->dest;
1549 fl6.saddr = iph->daddr;
1551 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1552 rt6_do_redirect(dst, NULL, skb);
1556 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1558 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1560 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1562 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1564 struct net_device *dev = dst->dev;
1565 unsigned int mtu = dst_mtu(dst);
1566 struct net *net = dev_net(dev);
1568 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1570 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1571 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1574 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1575 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1576 * IPV6_MAXPLEN is also valid and means: "any MSS,
1577 * rely only on pmtu discovery"
1579 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1584 static unsigned int ip6_mtu(const struct dst_entry *dst)
1586 const struct rt6_info *rt = (const struct rt6_info *)dst;
1587 unsigned int mtu = rt->rt6i_pmtu;
1588 struct inet6_dev *idev;
1593 mtu = dst_metric_raw(dst, RTAX_MTU);
1600 idev = __in6_dev_get(dst->dev);
1602 mtu = idev->cnf.mtu6;
1606 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1609 static struct dst_entry *icmp6_dst_gc_list;
1610 static DEFINE_SPINLOCK(icmp6_dst_lock);
1612 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1615 struct dst_entry *dst;
1616 struct rt6_info *rt;
1617 struct inet6_dev *idev = in6_dev_get(dev);
1618 struct net *net = dev_net(dev);
1620 if (unlikely(!idev))
1621 return ERR_PTR(-ENODEV);
1623 rt = ip6_dst_alloc(net, dev, 0);
1624 if (unlikely(!rt)) {
1626 dst = ERR_PTR(-ENOMEM);
1630 rt->dst.flags |= DST_HOST;
1631 rt->dst.output = ip6_output;
1632 atomic_set(&rt->dst.__refcnt, 1);
1633 rt->rt6i_gateway = fl6->daddr;
1634 rt->rt6i_dst.addr = fl6->daddr;
1635 rt->rt6i_dst.plen = 128;
1636 rt->rt6i_idev = idev;
1637 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1639 spin_lock_bh(&icmp6_dst_lock);
1640 rt->dst.next = icmp6_dst_gc_list;
1641 icmp6_dst_gc_list = &rt->dst;
1642 spin_unlock_bh(&icmp6_dst_lock);
1644 fib6_force_start_gc(net);
1646 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1652 int icmp6_dst_gc(void)
1654 struct dst_entry *dst, **pprev;
1657 spin_lock_bh(&icmp6_dst_lock);
1658 pprev = &icmp6_dst_gc_list;
1660 while ((dst = *pprev) != NULL) {
1661 if (!atomic_read(&dst->__refcnt)) {
1670 spin_unlock_bh(&icmp6_dst_lock);
1675 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1678 struct dst_entry *dst, **pprev;
1680 spin_lock_bh(&icmp6_dst_lock);
1681 pprev = &icmp6_dst_gc_list;
1682 while ((dst = *pprev) != NULL) {
1683 struct rt6_info *rt = (struct rt6_info *) dst;
1684 if (func(rt, arg)) {
1691 spin_unlock_bh(&icmp6_dst_lock);
1694 static int ip6_dst_gc(struct dst_ops *ops)
1696 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1697 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1698 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1699 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1700 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1701 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1704 entries = dst_entries_get_fast(ops);
1705 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1706 entries <= rt_max_size)
1709 net->ipv6.ip6_rt_gc_expire++;
1710 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1711 entries = dst_entries_get_slow(ops);
1712 if (entries < ops->gc_thresh)
1713 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1715 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1716 return entries > rt_max_size;
1719 static int ip6_convert_metrics(struct mx6_config *mxc,
1720 const struct fib6_config *cfg)
1722 bool ecn_ca = false;
1730 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1734 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1735 int type = nla_type(nla);
1740 if (unlikely(type > RTAX_MAX))
1743 if (type == RTAX_CC_ALGO) {
1744 char tmp[TCP_CA_NAME_MAX];
1746 nla_strlcpy(tmp, nla, sizeof(tmp));
1747 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1748 if (val == TCP_CA_UNSPEC)
1751 val = nla_get_u32(nla);
1753 if (type == RTAX_HOPLIMIT && val > 255)
1755 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1759 __set_bit(type - 1, mxc->mx_valid);
1763 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1764 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1774 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1776 struct net *net = cfg->fc_nlinfo.nl_net;
1777 struct rt6_info *rt = NULL;
1778 struct net_device *dev = NULL;
1779 struct inet6_dev *idev = NULL;
1780 struct fib6_table *table;
1784 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1786 #ifndef CONFIG_IPV6_SUBTREES
1787 if (cfg->fc_src_len)
1790 if (cfg->fc_ifindex) {
1792 dev = dev_get_by_index(net, cfg->fc_ifindex);
1795 idev = in6_dev_get(dev);
1800 if (cfg->fc_metric == 0)
1801 cfg->fc_metric = IP6_RT_PRIO_USER;
1804 if (cfg->fc_nlinfo.nlh &&
1805 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1806 table = fib6_get_table(net, cfg->fc_table);
1808 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1809 table = fib6_new_table(net, cfg->fc_table);
1812 table = fib6_new_table(net, cfg->fc_table);
1818 rt = ip6_dst_alloc(net, NULL,
1819 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1826 if (cfg->fc_flags & RTF_EXPIRES)
1827 rt6_set_expires(rt, jiffies +
1828 clock_t_to_jiffies(cfg->fc_expires));
1830 rt6_clean_expires(rt);
1832 if (cfg->fc_protocol == RTPROT_UNSPEC)
1833 cfg->fc_protocol = RTPROT_BOOT;
1834 rt->rt6i_protocol = cfg->fc_protocol;
1836 addr_type = ipv6_addr_type(&cfg->fc_dst);
1838 if (addr_type & IPV6_ADDR_MULTICAST)
1839 rt->dst.input = ip6_mc_input;
1840 else if (cfg->fc_flags & RTF_LOCAL)
1841 rt->dst.input = ip6_input;
1843 rt->dst.input = ip6_forward;
1845 rt->dst.output = ip6_output;
1847 if (cfg->fc_encap) {
1848 struct lwtunnel_state *lwtstate;
1850 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1851 cfg->fc_encap, AF_INET6, cfg,
1855 rt->dst.lwtstate = lwtstate_get(lwtstate);
1856 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1857 rt->dst.lwtstate->orig_output = rt->dst.output;
1858 rt->dst.output = lwtunnel_output;
1860 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1861 rt->dst.lwtstate->orig_input = rt->dst.input;
1862 rt->dst.input = lwtunnel_input;
1866 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1867 rt->rt6i_dst.plen = cfg->fc_dst_len;
1868 if (rt->rt6i_dst.plen == 128)
1869 rt->dst.flags |= DST_HOST;
1871 #ifdef CONFIG_IPV6_SUBTREES
1872 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1873 rt->rt6i_src.plen = cfg->fc_src_len;
1876 rt->rt6i_metric = cfg->fc_metric;
1878 /* We cannot add true routes via loopback here,
1879 they would result in kernel looping; promote them to reject routes
1881 if ((cfg->fc_flags & RTF_REJECT) ||
1882 (dev && (dev->flags & IFF_LOOPBACK) &&
1883 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1884 !(cfg->fc_flags & RTF_LOCAL))) {
1885 /* hold loopback dev/idev if we haven't done so. */
1886 if (dev != net->loopback_dev) {
1891 dev = net->loopback_dev;
1893 idev = in6_dev_get(dev);
1899 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1900 switch (cfg->fc_type) {
1902 rt->dst.error = -EINVAL;
1903 rt->dst.output = dst_discard_out;
1904 rt->dst.input = dst_discard;
1907 rt->dst.error = -EACCES;
1908 rt->dst.output = ip6_pkt_prohibit_out;
1909 rt->dst.input = ip6_pkt_prohibit;
1912 case RTN_UNREACHABLE:
1914 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1915 : (cfg->fc_type == RTN_UNREACHABLE)
1916 ? -EHOSTUNREACH : -ENETUNREACH;
1917 rt->dst.output = ip6_pkt_discard_out;
1918 rt->dst.input = ip6_pkt_discard;
1924 if (cfg->fc_flags & RTF_GATEWAY) {
1925 const struct in6_addr *gw_addr;
1928 gw_addr = &cfg->fc_gateway;
1929 gwa_type = ipv6_addr_type(gw_addr);
1931 /* if gw_addr is local we will fail to detect this in case
1932 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1933 * will return already-added prefix route via interface that
1934 * prefix route was assigned to, which might be non-loopback.
1937 if (ipv6_chk_addr_and_flags(net, gw_addr,
1938 gwa_type & IPV6_ADDR_LINKLOCAL ?
1942 rt->rt6i_gateway = *gw_addr;
1944 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1945 struct rt6_info *grt;
1947 /* IPv6 strictly inhibits using not link-local
1948 addresses as nexthop address.
1949 Otherwise, router will not able to send redirects.
1950 It is very good, but in some (rare!) circumstances
1951 (SIT, PtP, NBMA NOARP links) it is handy to allow
1952 some exceptions. --ANK
1954 if (!(gwa_type & IPV6_ADDR_UNICAST))
1957 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1959 err = -EHOSTUNREACH;
1963 if (dev != grt->dst.dev) {
1969 idev = grt->rt6i_idev;
1971 in6_dev_hold(grt->rt6i_idev);
1973 if (!(grt->rt6i_flags & RTF_GATEWAY))
1981 if (!dev || (dev->flags & IFF_LOOPBACK))
1989 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1990 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1994 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1995 rt->rt6i_prefsrc.plen = 128;
1997 rt->rt6i_prefsrc.plen = 0;
1999 rt->rt6i_flags = cfg->fc_flags;
2003 rt->rt6i_idev = idev;
2004 rt->rt6i_table = table;
2006 cfg->fc_nlinfo.nl_net = dev_net(dev);
2017 return ERR_PTR(err);
2020 int ip6_route_add(struct fib6_config *cfg)
2022 struct mx6_config mxc = { .mx = NULL, };
2023 struct rt6_info *rt;
2026 rt = ip6_route_info_create(cfg);
2033 err = ip6_convert_metrics(&mxc, cfg);
2037 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2049 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2052 struct fib6_table *table;
2053 struct net *net = dev_net(rt->dst.dev);
2055 if (rt == net->ipv6.ip6_null_entry ||
2056 rt->dst.flags & DST_NOCACHE) {
2061 table = rt->rt6i_table;
2062 write_lock_bh(&table->tb6_lock);
2063 err = fib6_del(rt, info);
2064 write_unlock_bh(&table->tb6_lock);
2071 int ip6_del_rt(struct rt6_info *rt)
2073 struct nl_info info = {
2074 .nl_net = dev_net(rt->dst.dev),
2076 return __ip6_del_rt(rt, &info);
2079 static int ip6_route_del(struct fib6_config *cfg)
2081 struct fib6_table *table;
2082 struct fib6_node *fn;
2083 struct rt6_info *rt;
2086 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2090 read_lock_bh(&table->tb6_lock);
2092 fn = fib6_locate(&table->tb6_root,
2093 &cfg->fc_dst, cfg->fc_dst_len,
2094 &cfg->fc_src, cfg->fc_src_len);
2097 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2098 if ((rt->rt6i_flags & RTF_CACHE) &&
2099 !(cfg->fc_flags & RTF_CACHE))
2101 if (cfg->fc_ifindex &&
2103 rt->dst.dev->ifindex != cfg->fc_ifindex))
2105 if (cfg->fc_flags & RTF_GATEWAY &&
2106 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2108 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2111 read_unlock_bh(&table->tb6_lock);
2113 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2116 read_unlock_bh(&table->tb6_lock);
2121 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2123 struct netevent_redirect netevent;
2124 struct rt6_info *rt, *nrt = NULL;
2125 struct ndisc_options ndopts;
2126 struct inet6_dev *in6_dev;
2127 struct neighbour *neigh;
2129 int optlen, on_link;
2132 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2133 optlen -= sizeof(*msg);
2136 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2140 msg = (struct rd_msg *)icmp6_hdr(skb);
2142 if (ipv6_addr_is_multicast(&msg->dest)) {
2143 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2148 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2150 } else if (ipv6_addr_type(&msg->target) !=
2151 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2152 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2156 in6_dev = __in6_dev_get(skb->dev);
2159 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2163 * The IP source address of the Redirect MUST be the same as the current
2164 * first-hop router for the specified ICMP Destination Address.
2167 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2168 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2173 if (ndopts.nd_opts_tgt_lladdr) {
2174 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2177 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2182 rt = (struct rt6_info *) dst;
2183 if (rt->rt6i_flags & RTF_REJECT) {
2184 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2188 /* Redirect received -> path was valid.
2189 * Look, redirects are sent only in response to data packets,
2190 * so that this nexthop apparently is reachable. --ANK
2192 dst_confirm(&rt->dst);
2194 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2199 * We have finally decided to accept it.
2202 neigh_update(neigh, lladdr, NUD_STALE,
2203 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2204 NEIGH_UPDATE_F_OVERRIDE|
2205 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2206 NEIGH_UPDATE_F_ISROUTER))
2209 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2213 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2215 nrt->rt6i_flags &= ~RTF_GATEWAY;
2217 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2219 if (ip6_ins_rt(nrt))
2222 netevent.old = &rt->dst;
2223 netevent.new = &nrt->dst;
2224 netevent.daddr = &msg->dest;
2225 netevent.neigh = neigh;
2226 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2228 if (rt->rt6i_flags & RTF_CACHE) {
2229 rt = (struct rt6_info *) dst_clone(&rt->dst);
2234 neigh_release(neigh);
2238 * Misc support functions
2241 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2243 BUG_ON(from->dst.from);
2245 rt->rt6i_flags &= ~RTF_EXPIRES;
2246 dst_hold(&from->dst);
2247 rt->dst.from = &from->dst;
2248 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2251 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2253 rt->dst.input = ort->dst.input;
2254 rt->dst.output = ort->dst.output;
2255 rt->rt6i_dst = ort->rt6i_dst;
2256 rt->dst.error = ort->dst.error;
2257 rt->rt6i_idev = ort->rt6i_idev;
2259 in6_dev_hold(rt->rt6i_idev);
2260 rt->dst.lastuse = jiffies;
2261 rt->rt6i_gateway = ort->rt6i_gateway;
2262 rt->rt6i_flags = ort->rt6i_flags;
2263 rt6_set_from(rt, ort);
2264 rt->rt6i_metric = ort->rt6i_metric;
2265 #ifdef CONFIG_IPV6_SUBTREES
2266 rt->rt6i_src = ort->rt6i_src;
2268 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2269 rt->rt6i_table = ort->rt6i_table;
2270 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2273 #ifdef CONFIG_IPV6_ROUTE_INFO
2274 static struct rt6_info *rt6_get_route_info(struct net *net,
2275 const struct in6_addr *prefix, int prefixlen,
2276 const struct in6_addr *gwaddr, int ifindex)
2278 struct fib6_node *fn;
2279 struct rt6_info *rt = NULL;
2280 struct fib6_table *table;
2282 table = fib6_get_table(net, RT6_TABLE_INFO);
2286 read_lock_bh(&table->tb6_lock);
2287 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2291 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2292 if (rt->dst.dev->ifindex != ifindex)
2294 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2296 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2302 read_unlock_bh(&table->tb6_lock);
2306 static struct rt6_info *rt6_add_route_info(struct net *net,
2307 const struct in6_addr *prefix, int prefixlen,
2308 const struct in6_addr *gwaddr, int ifindex,
2311 struct fib6_config cfg = {
2312 .fc_metric = IP6_RT_PRIO_USER,
2313 .fc_ifindex = ifindex,
2314 .fc_dst_len = prefixlen,
2315 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2316 RTF_UP | RTF_PREF(pref),
2317 .fc_nlinfo.portid = 0,
2318 .fc_nlinfo.nlh = NULL,
2319 .fc_nlinfo.nl_net = net,
2322 cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2323 cfg.fc_dst = *prefix;
2324 cfg.fc_gateway = *gwaddr;
2326 /* We should treat it as a default route if prefix length is 0. */
2328 cfg.fc_flags |= RTF_DEFAULT;
2330 ip6_route_add(&cfg);
2332 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2336 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2338 struct rt6_info *rt;
2339 struct fib6_table *table;
2341 table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2345 read_lock_bh(&table->tb6_lock);
2346 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2347 if (dev == rt->dst.dev &&
2348 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2349 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2354 read_unlock_bh(&table->tb6_lock);
2358 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2359 struct net_device *dev,
2362 struct fib6_config cfg = {
2363 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2364 .fc_metric = IP6_RT_PRIO_USER,
2365 .fc_ifindex = dev->ifindex,
2366 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2367 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2368 .fc_nlinfo.portid = 0,
2369 .fc_nlinfo.nlh = NULL,
2370 .fc_nlinfo.nl_net = dev_net(dev),
2373 cfg.fc_gateway = *gwaddr;
2375 ip6_route_add(&cfg);
2377 return rt6_get_dflt_router(gwaddr, dev);
2380 void rt6_purge_dflt_routers(struct net *net)
2382 struct rt6_info *rt;
2383 struct fib6_table *table;
2385 /* NOTE: Keep consistent with rt6_get_dflt_router */
2386 table = fib6_get_table(net, RT6_TABLE_DFLT);
2391 read_lock_bh(&table->tb6_lock);
2392 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2393 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2394 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2396 read_unlock_bh(&table->tb6_lock);
2401 read_unlock_bh(&table->tb6_lock);
2404 static void rtmsg_to_fib6_config(struct net *net,
2405 struct in6_rtmsg *rtmsg,
2406 struct fib6_config *cfg)
2408 memset(cfg, 0, sizeof(*cfg));
2410 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2412 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2413 cfg->fc_metric = rtmsg->rtmsg_metric;
2414 cfg->fc_expires = rtmsg->rtmsg_info;
2415 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2416 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2417 cfg->fc_flags = rtmsg->rtmsg_flags;
2419 cfg->fc_nlinfo.nl_net = net;
2421 cfg->fc_dst = rtmsg->rtmsg_dst;
2422 cfg->fc_src = rtmsg->rtmsg_src;
2423 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2426 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2428 struct fib6_config cfg;
2429 struct in6_rtmsg rtmsg;
2433 case SIOCADDRT: /* Add a route */
2434 case SIOCDELRT: /* Delete a route */
2435 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2437 err = copy_from_user(&rtmsg, arg,
2438 sizeof(struct in6_rtmsg));
2442 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2447 err = ip6_route_add(&cfg);
2450 err = ip6_route_del(&cfg);
2464 * Drop the packet on the floor
2467 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2470 struct dst_entry *dst = skb_dst(skb);
2471 switch (ipstats_mib_noroutes) {
2472 case IPSTATS_MIB_INNOROUTES:
2473 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2474 if (type == IPV6_ADDR_ANY) {
2475 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2476 IPSTATS_MIB_INADDRERRORS);
2480 case IPSTATS_MIB_OUTNOROUTES:
2481 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2482 ipstats_mib_noroutes);
2485 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2490 static int ip6_pkt_discard(struct sk_buff *skb)
2492 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2495 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2497 skb->dev = skb_dst(skb)->dev;
2498 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2501 static int ip6_pkt_prohibit(struct sk_buff *skb)
2503 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2506 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2508 skb->dev = skb_dst(skb)->dev;
2509 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2513 * Allocate a dst for local (unicast / anycast) address.
2516 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2517 const struct in6_addr *addr,
2521 struct net *net = dev_net(idev->dev);
2522 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2525 return ERR_PTR(-ENOMEM);
2529 rt->dst.flags |= DST_HOST;
2530 rt->dst.input = ip6_input;
2531 rt->dst.output = ip6_output;
2532 rt->rt6i_idev = idev;
2534 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2536 rt->rt6i_flags |= RTF_ANYCAST;
2538 rt->rt6i_flags |= RTF_LOCAL;
2540 rt->rt6i_gateway = *addr;
2541 rt->rt6i_dst.addr = *addr;
2542 rt->rt6i_dst.plen = 128;
2543 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2544 rt->rt6i_table = fib6_get_table(net, tb_id);
2545 rt->dst.flags |= DST_NOCACHE;
2547 atomic_set(&rt->dst.__refcnt, 1);
2552 int ip6_route_get_saddr(struct net *net,
2553 struct rt6_info *rt,
2554 const struct in6_addr *daddr,
2556 struct in6_addr *saddr)
2558 struct inet6_dev *idev =
2559 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2561 if (rt && rt->rt6i_prefsrc.plen)
2562 *saddr = rt->rt6i_prefsrc.addr;
2564 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2565 daddr, prefs, saddr);
2569 /* remove deleted ip from prefsrc entries */
2570 struct arg_dev_net_ip {
2571 struct net_device *dev;
2573 struct in6_addr *addr;
2576 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2578 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2579 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2580 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2582 if (((void *)rt->dst.dev == dev || !dev) &&
2583 rt != net->ipv6.ip6_null_entry &&
2584 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2585 /* remove prefsrc entry */
2586 rt->rt6i_prefsrc.plen = 0;
2591 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2593 struct net *net = dev_net(ifp->idev->dev);
2594 struct arg_dev_net_ip adni = {
2595 .dev = ifp->idev->dev,
2599 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2602 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2603 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2605 /* Remove routers and update dst entries when gateway turn into host. */
2606 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2608 struct in6_addr *gateway = (struct in6_addr *)arg;
2610 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2611 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2612 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2618 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2620 fib6_clean_all(net, fib6_clean_tohost, gateway);
2623 struct arg_dev_net {
2624 struct net_device *dev;
2628 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2630 const struct arg_dev_net *adn = arg;
2631 const struct net_device *dev = adn->dev;
2633 if ((rt->dst.dev == dev || !dev) &&
2634 rt != adn->net->ipv6.ip6_null_entry)
2640 void rt6_ifdown(struct net *net, struct net_device *dev)
2642 struct arg_dev_net adn = {
2647 fib6_clean_all(net, fib6_ifdown, &adn);
2648 icmp6_clean_all(fib6_ifdown, &adn);
2650 rt6_uncached_list_flush_dev(net, dev);
2653 struct rt6_mtu_change_arg {
2654 struct net_device *dev;
2658 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2660 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2661 struct inet6_dev *idev;
2663 /* In IPv6 pmtu discovery is not optional,
2664 so that RTAX_MTU lock cannot disable it.
2665 We still use this lock to block changes
2666 caused by addrconf/ndisc.
2669 idev = __in6_dev_get(arg->dev);
2673 /* For administrative MTU increase, there is no way to discover
2674 IPv6 PMTU increase, so PMTU increase should be updated here.
2675 Since RFC 1981 doesn't include administrative MTU increase
2676 update PMTU increase is a MUST. (i.e. jumbo frame)
2679 If new MTU is less than route PMTU, this new MTU will be the
2680 lowest MTU in the path, update the route PMTU to reflect PMTU
2681 decreases; if new MTU is greater than route PMTU, and the
2682 old MTU is the lowest MTU in the path, update the route PMTU
2683 to reflect the increase. In this case if the other nodes' MTU
2684 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2687 if (rt->dst.dev == arg->dev &&
2688 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2689 if (rt->rt6i_flags & RTF_CACHE) {
2690 /* For RTF_CACHE with rt6i_pmtu == 0
2691 * (i.e. a redirected route),
2692 * the metrics of its rt->dst.from has already
2695 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2696 rt->rt6i_pmtu = arg->mtu;
2697 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2698 (dst_mtu(&rt->dst) < arg->mtu &&
2699 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2700 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2706 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2708 struct rt6_mtu_change_arg arg = {
2713 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2716 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2717 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2718 [RTA_OIF] = { .type = NLA_U32 },
2719 [RTA_IIF] = { .type = NLA_U32 },
2720 [RTA_PRIORITY] = { .type = NLA_U32 },
2721 [RTA_METRICS] = { .type = NLA_NESTED },
2722 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2723 [RTA_PREF] = { .type = NLA_U8 },
2724 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2725 [RTA_ENCAP] = { .type = NLA_NESTED },
2726 [RTA_EXPIRES] = { .type = NLA_U32 },
2729 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2730 struct fib6_config *cfg)
2733 struct nlattr *tb[RTA_MAX+1];
2737 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2742 rtm = nlmsg_data(nlh);
2743 memset(cfg, 0, sizeof(*cfg));
2745 cfg->fc_table = rtm->rtm_table;
2746 cfg->fc_dst_len = rtm->rtm_dst_len;
2747 cfg->fc_src_len = rtm->rtm_src_len;
2748 cfg->fc_flags = RTF_UP;
2749 cfg->fc_protocol = rtm->rtm_protocol;
2750 cfg->fc_type = rtm->rtm_type;
2752 if (rtm->rtm_type == RTN_UNREACHABLE ||
2753 rtm->rtm_type == RTN_BLACKHOLE ||
2754 rtm->rtm_type == RTN_PROHIBIT ||
2755 rtm->rtm_type == RTN_THROW)
2756 cfg->fc_flags |= RTF_REJECT;
2758 if (rtm->rtm_type == RTN_LOCAL)
2759 cfg->fc_flags |= RTF_LOCAL;
2761 if (rtm->rtm_flags & RTM_F_CLONED)
2762 cfg->fc_flags |= RTF_CACHE;
2764 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2765 cfg->fc_nlinfo.nlh = nlh;
2766 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2768 if (tb[RTA_GATEWAY]) {
2769 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2770 cfg->fc_flags |= RTF_GATEWAY;
2774 int plen = (rtm->rtm_dst_len + 7) >> 3;
2776 if (nla_len(tb[RTA_DST]) < plen)
2779 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2783 int plen = (rtm->rtm_src_len + 7) >> 3;
2785 if (nla_len(tb[RTA_SRC]) < plen)
2788 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2791 if (tb[RTA_PREFSRC])
2792 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2795 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2797 if (tb[RTA_PRIORITY])
2798 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2800 if (tb[RTA_METRICS]) {
2801 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2802 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2806 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2808 if (tb[RTA_MULTIPATH]) {
2809 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2810 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2814 pref = nla_get_u8(tb[RTA_PREF]);
2815 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2816 pref != ICMPV6_ROUTER_PREF_HIGH)
2817 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2818 cfg->fc_flags |= RTF_PREF(pref);
2822 cfg->fc_encap = tb[RTA_ENCAP];
2824 if (tb[RTA_ENCAP_TYPE])
2825 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2827 if (tb[RTA_EXPIRES]) {
2828 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2830 if (addrconf_finite_timeout(timeout)) {
2831 cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2832 cfg->fc_flags |= RTF_EXPIRES;
2842 struct rt6_info *rt6_info;
2843 struct fib6_config r_cfg;
2844 struct mx6_config mxc;
2845 struct list_head next;
2848 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2852 list_for_each_entry(nh, rt6_nh_list, next) {
2853 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2854 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2855 nh->r_cfg.fc_ifindex);
2859 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2860 struct rt6_info *rt, struct fib6_config *r_cfg)
2863 struct rt6_info *rtnh;
2866 list_for_each_entry(nh, rt6_nh_list, next) {
2867 /* check if rt6_info already exists */
2868 rtnh = nh->rt6_info;
2870 if (rtnh->dst.dev == rt->dst.dev &&
2871 rtnh->rt6i_idev == rt->rt6i_idev &&
2872 ipv6_addr_equal(&rtnh->rt6i_gateway,
2877 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2881 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2886 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2887 list_add_tail(&nh->next, rt6_nh_list);
2892 static int ip6_route_multipath_add(struct fib6_config *cfg)
2894 struct fib6_config r_cfg;
2895 struct rtnexthop *rtnh;
2896 struct rt6_info *rt;
2897 struct rt6_nh *err_nh;
2898 struct rt6_nh *nh, *nh_safe;
2903 int replace = (cfg->fc_nlinfo.nlh &&
2904 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2905 LIST_HEAD(rt6_nh_list);
2907 remaining = cfg->fc_mp_len;
2908 rtnh = (struct rtnexthop *)cfg->fc_mp;
2910 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2911 * rt6_info structs per nexthop
2913 while (rtnh_ok(rtnh, remaining)) {
2914 memcpy(&r_cfg, cfg, sizeof(*cfg));
2915 if (rtnh->rtnh_ifindex)
2916 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2918 attrlen = rtnh_attrlen(rtnh);
2920 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2922 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2924 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2925 r_cfg.fc_flags |= RTF_GATEWAY;
2927 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2928 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2930 r_cfg.fc_encap_type = nla_get_u16(nla);
2933 rt = ip6_route_info_create(&r_cfg);
2940 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2946 rtnh = rtnh_next(rtnh, &remaining);
2950 list_for_each_entry(nh, &rt6_nh_list, next) {
2951 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2952 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2953 nh->rt6_info = NULL;
2956 ip6_print_replace_route_err(&rt6_nh_list);
2961 /* Because each route is added like a single route we remove
2962 * these flags after the first nexthop: if there is a collision,
2963 * we have already failed to add the first nexthop:
2964 * fib6_add_rt2node() has rejected it; when replacing, old
2965 * nexthops have been replaced by first new, the rest should
2968 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2976 /* Delete routes that were already added */
2977 list_for_each_entry(nh, &rt6_nh_list, next) {
2980 ip6_route_del(&nh->r_cfg);
2984 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2986 dst_free(&nh->rt6_info->dst);
2988 list_del(&nh->next);
2995 static int ip6_route_multipath_del(struct fib6_config *cfg)
2997 struct fib6_config r_cfg;
2998 struct rtnexthop *rtnh;
3001 int err = 1, last_err = 0;
3003 remaining = cfg->fc_mp_len;
3004 rtnh = (struct rtnexthop *)cfg->fc_mp;
3006 /* Parse a Multipath Entry */
3007 while (rtnh_ok(rtnh, remaining)) {
3008 memcpy(&r_cfg, cfg, sizeof(*cfg));
3009 if (rtnh->rtnh_ifindex)
3010 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3012 attrlen = rtnh_attrlen(rtnh);
3014 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3016 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3018 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3019 r_cfg.fc_flags |= RTF_GATEWAY;
3022 err = ip6_route_del(&r_cfg);
3026 rtnh = rtnh_next(rtnh, &remaining);
3032 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3034 struct fib6_config cfg;
3037 err = rtm_to_fib6_config(skb, nlh, &cfg);
3042 return ip6_route_multipath_del(&cfg);
3044 return ip6_route_del(&cfg);
3047 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3049 struct fib6_config cfg;
3052 err = rtm_to_fib6_config(skb, nlh, &cfg);
3057 return ip6_route_multipath_add(&cfg);
3059 return ip6_route_add(&cfg);
3062 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3064 return NLMSG_ALIGN(sizeof(struct rtmsg))
3065 + nla_total_size(16) /* RTA_SRC */
3066 + nla_total_size(16) /* RTA_DST */
3067 + nla_total_size(16) /* RTA_GATEWAY */
3068 + nla_total_size(16) /* RTA_PREFSRC */
3069 + nla_total_size(4) /* RTA_TABLE */
3070 + nla_total_size(4) /* RTA_IIF */
3071 + nla_total_size(4) /* RTA_OIF */
3072 + nla_total_size(4) /* RTA_PRIORITY */
3073 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3074 + nla_total_size(sizeof(struct rta_cacheinfo))
3075 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3076 + nla_total_size(1) /* RTA_PREF */
3077 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3080 static int rt6_fill_node(struct net *net,
3081 struct sk_buff *skb, struct rt6_info *rt,
3082 struct in6_addr *dst, struct in6_addr *src,
3083 int iif, int type, u32 portid, u32 seq,
3084 int prefix, int nowait, unsigned int flags)
3086 u32 metrics[RTAX_MAX];
3088 struct nlmsghdr *nlh;
3092 if (prefix) { /* user wants prefix routes only */
3093 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3094 /* success since this is not a prefix route */
3099 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3103 rtm = nlmsg_data(nlh);
3104 rtm->rtm_family = AF_INET6;
3105 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3106 rtm->rtm_src_len = rt->rt6i_src.plen;
3109 table = rt->rt6i_table->tb6_id;
3111 table = RT6_TABLE_UNSPEC;
3112 rtm->rtm_table = table;
3113 if (nla_put_u32(skb, RTA_TABLE, table))
3114 goto nla_put_failure;
3115 if (rt->rt6i_flags & RTF_REJECT) {
3116 switch (rt->dst.error) {
3118 rtm->rtm_type = RTN_BLACKHOLE;
3121 rtm->rtm_type = RTN_PROHIBIT;
3124 rtm->rtm_type = RTN_THROW;
3127 rtm->rtm_type = RTN_UNREACHABLE;
3131 else if (rt->rt6i_flags & RTF_LOCAL)
3132 rtm->rtm_type = RTN_LOCAL;
3133 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3134 rtm->rtm_type = RTN_LOCAL;
3136 rtm->rtm_type = RTN_UNICAST;
3138 if (!netif_carrier_ok(rt->dst.dev)) {
3139 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3140 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3141 rtm->rtm_flags |= RTNH_F_DEAD;
3143 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3144 rtm->rtm_protocol = rt->rt6i_protocol;
3145 if (rt->rt6i_flags & RTF_DYNAMIC)
3146 rtm->rtm_protocol = RTPROT_REDIRECT;
3147 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3148 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3149 rtm->rtm_protocol = RTPROT_RA;
3151 rtm->rtm_protocol = RTPROT_KERNEL;
3154 if (rt->rt6i_flags & RTF_CACHE)
3155 rtm->rtm_flags |= RTM_F_CLONED;
3158 if (nla_put_in6_addr(skb, RTA_DST, dst))
3159 goto nla_put_failure;
3160 rtm->rtm_dst_len = 128;
3161 } else if (rtm->rtm_dst_len)
3162 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3163 goto nla_put_failure;
3164 #ifdef CONFIG_IPV6_SUBTREES
3166 if (nla_put_in6_addr(skb, RTA_SRC, src))
3167 goto nla_put_failure;
3168 rtm->rtm_src_len = 128;
3169 } else if (rtm->rtm_src_len &&
3170 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3171 goto nla_put_failure;
3174 #ifdef CONFIG_IPV6_MROUTE
3175 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3176 int err = ip6mr_get_route(net, skb, rtm, nowait);
3181 goto nla_put_failure;
3183 if (err == -EMSGSIZE)
3184 goto nla_put_failure;
3189 if (nla_put_u32(skb, RTA_IIF, iif))
3190 goto nla_put_failure;
3192 struct in6_addr saddr_buf;
3193 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3194 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3195 goto nla_put_failure;
3198 if (rt->rt6i_prefsrc.plen) {
3199 struct in6_addr saddr_buf;
3200 saddr_buf = rt->rt6i_prefsrc.addr;
3201 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3202 goto nla_put_failure;
3205 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3207 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3208 if (rtnetlink_put_metrics(skb, metrics) < 0)
3209 goto nla_put_failure;
3211 if (rt->rt6i_flags & RTF_GATEWAY) {
3212 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3213 goto nla_put_failure;
3217 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3218 goto nla_put_failure;
3219 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3220 goto nla_put_failure;
3222 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3224 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3225 goto nla_put_failure;
3227 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3228 goto nla_put_failure;
3230 lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3232 nlmsg_end(skb, nlh);
3236 nlmsg_cancel(skb, nlh);
3240 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3242 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3245 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3246 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3247 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3251 return rt6_fill_node(arg->net,
3252 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3253 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3254 prefix, 0, NLM_F_MULTI);
3257 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3259 struct net *net = sock_net(in_skb->sk);
3260 struct nlattr *tb[RTA_MAX+1];
3261 struct rt6_info *rt;
3262 struct sk_buff *skb;
3265 int err, iif = 0, oif = 0;
3267 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3272 memset(&fl6, 0, sizeof(fl6));
3275 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3278 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3282 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3285 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3289 iif = nla_get_u32(tb[RTA_IIF]);
3292 oif = nla_get_u32(tb[RTA_OIF]);
3295 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3298 struct net_device *dev;
3301 dev = __dev_get_by_index(net, iif);
3307 fl6.flowi6_iif = iif;
3309 if (!ipv6_addr_any(&fl6.saddr))
3310 flags |= RT6_LOOKUP_F_HAS_SADDR;
3312 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3315 fl6.flowi6_oif = oif;
3317 if (netif_index_is_l3_master(net, oif)) {
3318 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3319 FLOWI_FLAG_SKIP_NH_OIF;
3322 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3325 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3332 /* Reserve room for dummy headers, this skb can pass
3333 through good chunk of routing engine.
3335 skb_reset_mac_header(skb);
3336 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3338 skb_dst_set(skb, &rt->dst);
3340 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3341 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3342 nlh->nlmsg_seq, 0, 0, 0);
3348 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3353 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3354 unsigned int nlm_flags)
3356 struct sk_buff *skb;
3357 struct net *net = info->nl_net;
3362 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3364 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3368 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3369 event, info->portid, seq, 0, 0, nlm_flags);
3371 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3372 WARN_ON(err == -EMSGSIZE);
3376 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3377 info->nlh, gfp_any());
3381 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3384 static int ip6_route_dev_notify(struct notifier_block *this,
3385 unsigned long event, void *ptr)
3387 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3388 struct net *net = dev_net(dev);
3390 if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3391 net->ipv6.ip6_null_entry->dst.dev = dev;
3392 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3393 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3394 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3395 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3396 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3397 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3408 #ifdef CONFIG_PROC_FS
3410 static const struct file_operations ipv6_route_proc_fops = {
3411 .owner = THIS_MODULE,
3412 .open = ipv6_route_open,
3414 .llseek = seq_lseek,
3415 .release = seq_release_net,
3418 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3420 struct net *net = (struct net *)seq->private;
3421 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3422 net->ipv6.rt6_stats->fib_nodes,
3423 net->ipv6.rt6_stats->fib_route_nodes,
3424 net->ipv6.rt6_stats->fib_rt_alloc,
3425 net->ipv6.rt6_stats->fib_rt_entries,
3426 net->ipv6.rt6_stats->fib_rt_cache,
3427 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3428 net->ipv6.rt6_stats->fib_discarded_routes);
3433 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3435 return single_open_net(inode, file, rt6_stats_seq_show);
3438 static const struct file_operations rt6_stats_seq_fops = {
3439 .owner = THIS_MODULE,
3440 .open = rt6_stats_seq_open,
3442 .llseek = seq_lseek,
3443 .release = single_release_net,
3445 #endif /* CONFIG_PROC_FS */
3447 #ifdef CONFIG_SYSCTL
3450 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3451 void __user *buffer, size_t *lenp, loff_t *ppos)
3458 net = (struct net *)ctl->extra1;
3459 delay = net->ipv6.sysctl.flush_delay;
3460 proc_dointvec(ctl, write, buffer, lenp, ppos);
3461 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3465 struct ctl_table ipv6_route_table_template[] = {
3467 .procname = "flush",
3468 .data = &init_net.ipv6.sysctl.flush_delay,
3469 .maxlen = sizeof(int),
3471 .proc_handler = ipv6_sysctl_rtcache_flush
3474 .procname = "gc_thresh",
3475 .data = &ip6_dst_ops_template.gc_thresh,
3476 .maxlen = sizeof(int),
3478 .proc_handler = proc_dointvec,
3481 .procname = "max_size",
3482 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3483 .maxlen = sizeof(int),
3485 .proc_handler = proc_dointvec,
3488 .procname = "gc_min_interval",
3489 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3490 .maxlen = sizeof(int),
3492 .proc_handler = proc_dointvec_jiffies,
3495 .procname = "gc_timeout",
3496 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3497 .maxlen = sizeof(int),
3499 .proc_handler = proc_dointvec_jiffies,
3502 .procname = "gc_interval",
3503 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3504 .maxlen = sizeof(int),
3506 .proc_handler = proc_dointvec_jiffies,
3509 .procname = "gc_elasticity",
3510 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3511 .maxlen = sizeof(int),
3513 .proc_handler = proc_dointvec,
3516 .procname = "mtu_expires",
3517 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3518 .maxlen = sizeof(int),
3520 .proc_handler = proc_dointvec_jiffies,
3523 .procname = "min_adv_mss",
3524 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3525 .maxlen = sizeof(int),
3527 .proc_handler = proc_dointvec,
3530 .procname = "gc_min_interval_ms",
3531 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3532 .maxlen = sizeof(int),
3534 .proc_handler = proc_dointvec_ms_jiffies,
3539 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3541 struct ctl_table *table;
3543 table = kmemdup(ipv6_route_table_template,
3544 sizeof(ipv6_route_table_template),
3548 table[0].data = &net->ipv6.sysctl.flush_delay;
3549 table[0].extra1 = net;
3550 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3551 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3552 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3553 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3554 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3555 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3556 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3557 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3558 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3560 /* Don't export sysctls to unprivileged users */
3561 if (net->user_ns != &init_user_ns)
3562 table[0].procname = NULL;
3569 static int __net_init ip6_route_net_init(struct net *net)
3573 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3574 sizeof(net->ipv6.ip6_dst_ops));
3576 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3577 goto out_ip6_dst_ops;
3579 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3580 sizeof(*net->ipv6.ip6_null_entry),
3582 if (!net->ipv6.ip6_null_entry)
3583 goto out_ip6_dst_entries;
3584 net->ipv6.ip6_null_entry->dst.path =
3585 (struct dst_entry *)net->ipv6.ip6_null_entry;
3586 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3587 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3588 ip6_template_metrics, true);
3590 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3591 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3592 sizeof(*net->ipv6.ip6_prohibit_entry),
3594 if (!net->ipv6.ip6_prohibit_entry)
3595 goto out_ip6_null_entry;
3596 net->ipv6.ip6_prohibit_entry->dst.path =
3597 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3598 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3599 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3600 ip6_template_metrics, true);
3602 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3603 sizeof(*net->ipv6.ip6_blk_hole_entry),
3605 if (!net->ipv6.ip6_blk_hole_entry)
3606 goto out_ip6_prohibit_entry;
3607 net->ipv6.ip6_blk_hole_entry->dst.path =
3608 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3609 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3610 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3611 ip6_template_metrics, true);
3614 net->ipv6.sysctl.flush_delay = 0;
3615 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3616 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3617 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3618 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3619 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3620 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3621 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3623 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3629 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3630 out_ip6_prohibit_entry:
3631 kfree(net->ipv6.ip6_prohibit_entry);
3633 kfree(net->ipv6.ip6_null_entry);
3635 out_ip6_dst_entries:
3636 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3641 static void __net_exit ip6_route_net_exit(struct net *net)
3643 kfree(net->ipv6.ip6_null_entry);
3644 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3645 kfree(net->ipv6.ip6_prohibit_entry);
3646 kfree(net->ipv6.ip6_blk_hole_entry);
3648 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3651 static int __net_init ip6_route_net_init_late(struct net *net)
3653 #ifdef CONFIG_PROC_FS
3654 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3655 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3660 static void __net_exit ip6_route_net_exit_late(struct net *net)
3662 #ifdef CONFIG_PROC_FS
3663 remove_proc_entry("ipv6_route", net->proc_net);
3664 remove_proc_entry("rt6_stats", net->proc_net);
3668 static struct pernet_operations ip6_route_net_ops = {
3669 .init = ip6_route_net_init,
3670 .exit = ip6_route_net_exit,
3673 static int __net_init ipv6_inetpeer_init(struct net *net)
3675 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3679 inet_peer_base_init(bp);
3680 net->ipv6.peers = bp;
3684 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3686 struct inet_peer_base *bp = net->ipv6.peers;
3688 net->ipv6.peers = NULL;
3689 inetpeer_invalidate_tree(bp);
3693 static struct pernet_operations ipv6_inetpeer_ops = {
3694 .init = ipv6_inetpeer_init,
3695 .exit = ipv6_inetpeer_exit,
3698 static struct pernet_operations ip6_route_net_late_ops = {
3699 .init = ip6_route_net_init_late,
3700 .exit = ip6_route_net_exit_late,
3703 static struct notifier_block ip6_route_dev_notifier = {
3704 .notifier_call = ip6_route_dev_notify,
3708 int __init ip6_route_init(void)
3714 ip6_dst_ops_template.kmem_cachep =
3715 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3716 SLAB_HWCACHE_ALIGN, NULL);
3717 if (!ip6_dst_ops_template.kmem_cachep)
3720 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3722 goto out_kmem_cache;
3724 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3726 goto out_dst_entries;
3728 ret = register_pernet_subsys(&ip6_route_net_ops);
3730 goto out_register_inetpeer;
3732 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3734 /* Registering of the loopback is done before this portion of code,
3735 * the loopback reference in rt6_info will not be taken, do it
3736 * manually for init_net */
3737 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3738 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3739 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3740 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3741 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3742 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3743 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3747 goto out_register_subsys;
3753 ret = fib6_rules_init();
3757 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3759 goto fib6_rules_init;
3762 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3763 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3764 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3765 goto out_register_late_subsys;
3767 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3769 goto out_register_late_subsys;
3771 for_each_possible_cpu(cpu) {
3772 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3774 INIT_LIST_HEAD(&ul->head);
3775 spin_lock_init(&ul->lock);
3781 out_register_late_subsys:
3782 unregister_pernet_subsys(&ip6_route_net_late_ops);
3784 fib6_rules_cleanup();
3789 out_register_subsys:
3790 unregister_pernet_subsys(&ip6_route_net_ops);
3791 out_register_inetpeer:
3792 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3794 dst_entries_destroy(&ip6_dst_blackhole_ops);
3796 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3800 void ip6_route_cleanup(void)
3802 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3803 unregister_pernet_subsys(&ip6_route_net_late_ops);
3804 fib6_rules_cleanup();
3807 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3808 unregister_pernet_subsys(&ip6_route_net_ops);
3809 dst_entries_destroy(&ip6_dst_blackhole_ops);
3810 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);