2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
57 #include <net/dst_metadata.h>
59 #include <net/netevent.h>
60 #include <net/netlink.h>
61 #include <net/nexthop.h>
62 #include <net/lwtunnel.h>
63 #include <net/ip_tunnels.h>
64 #include <net/l3mdev.h>
66 #include <asm/uaccess.h>
69 #include <linux/sysctl.h>
73 RT6_NUD_FAIL_HARD = -3,
74 RT6_NUD_FAIL_PROBE = -2,
75 RT6_NUD_FAIL_DO_RR = -1,
79 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
80 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
81 static unsigned int ip6_default_advmss(const struct dst_entry *dst);
82 static unsigned int ip6_mtu(const struct dst_entry *dst);
83 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
84 static void ip6_dst_destroy(struct dst_entry *);
85 static void ip6_dst_ifdown(struct dst_entry *,
86 struct net_device *dev, int how);
87 static int ip6_dst_gc(struct dst_ops *ops);
89 static int ip6_pkt_discard(struct sk_buff *skb);
90 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
91 static int ip6_pkt_prohibit(struct sk_buff *skb);
92 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
93 static void ip6_link_failure(struct sk_buff *skb);
94 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
95 struct sk_buff *skb, u32 mtu);
96 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
98 static void rt6_dst_from_metrics_check(struct rt6_info *rt);
99 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
101 #ifdef CONFIG_IPV6_ROUTE_INFO
102 static struct rt6_info *rt6_add_route_info(struct net *net,
103 const struct in6_addr *prefix, int prefixlen,
104 const struct in6_addr *gwaddr,
105 struct net_device *dev,
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,
110 struct net_device *dev);
113 struct uncached_list {
115 struct list_head head;
118 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
120 static void rt6_uncached_list_add(struct rt6_info *rt)
122 struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
124 rt->dst.flags |= DST_NOCACHE;
125 rt->rt6i_uncached_list = ul;
127 spin_lock_bh(&ul->lock);
128 list_add_tail(&rt->rt6i_uncached, &ul->head);
129 spin_unlock_bh(&ul->lock);
132 static void rt6_uncached_list_del(struct rt6_info *rt)
134 if (!list_empty(&rt->rt6i_uncached)) {
135 struct uncached_list *ul = rt->rt6i_uncached_list;
137 spin_lock_bh(&ul->lock);
138 list_del(&rt->rt6i_uncached);
139 spin_unlock_bh(&ul->lock);
143 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
145 struct net_device *loopback_dev = net->loopback_dev;
148 if (dev == loopback_dev)
151 for_each_possible_cpu(cpu) {
152 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
155 spin_lock_bh(&ul->lock);
156 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
157 struct inet6_dev *rt_idev = rt->rt6i_idev;
158 struct net_device *rt_dev = rt->dst.dev;
160 if (rt_idev->dev == dev) {
161 rt->rt6i_idev = in6_dev_get(loopback_dev);
162 in6_dev_put(rt_idev);
166 rt->dst.dev = loopback_dev;
167 dev_hold(rt->dst.dev);
171 spin_unlock_bh(&ul->lock);
175 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
177 return dst_metrics_write_ptr(rt->dst.from);
180 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
182 struct rt6_info *rt = (struct rt6_info *)dst;
184 if (rt->rt6i_flags & RTF_PCPU)
185 return rt6_pcpu_cow_metrics(rt);
186 else if (rt->rt6i_flags & RTF_CACHE)
189 return dst_cow_metrics_generic(dst, old);
192 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
196 struct in6_addr *p = &rt->rt6i_gateway;
198 if (!ipv6_addr_any(p))
199 return (const void *) p;
201 return &ipv6_hdr(skb)->daddr;
205 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
209 struct rt6_info *rt = (struct rt6_info *) dst;
212 daddr = choose_neigh_daddr(rt, skb, daddr);
213 n = __ipv6_neigh_lookup(dst->dev, daddr);
216 return neigh_create(&nd_tbl, daddr, dst->dev);
219 static struct dst_ops ip6_dst_ops_template = {
223 .check = ip6_dst_check,
224 .default_advmss = ip6_default_advmss,
226 .cow_metrics = ipv6_cow_metrics,
227 .destroy = ip6_dst_destroy,
228 .ifdown = ip6_dst_ifdown,
229 .negative_advice = ip6_negative_advice,
230 .link_failure = ip6_link_failure,
231 .update_pmtu = ip6_rt_update_pmtu,
232 .redirect = rt6_do_redirect,
233 .local_out = __ip6_local_out,
234 .neigh_lookup = ip6_neigh_lookup,
237 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
239 unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
241 return mtu ? : dst->dev->mtu;
244 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
245 struct sk_buff *skb, u32 mtu)
249 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
254 static struct dst_ops ip6_dst_blackhole_ops = {
256 .destroy = ip6_dst_destroy,
257 .check = ip6_dst_check,
258 .mtu = ip6_blackhole_mtu,
259 .default_advmss = ip6_default_advmss,
260 .update_pmtu = ip6_rt_blackhole_update_pmtu,
261 .redirect = ip6_rt_blackhole_redirect,
262 .cow_metrics = dst_cow_metrics_generic,
263 .neigh_lookup = ip6_neigh_lookup,
266 static const u32 ip6_template_metrics[RTAX_MAX] = {
267 [RTAX_HOPLIMIT - 1] = 0,
270 static const struct rt6_info ip6_null_entry_template = {
272 .__refcnt = ATOMIC_INIT(1),
274 .obsolete = DST_OBSOLETE_FORCE_CHK,
275 .error = -ENETUNREACH,
276 .input = ip6_pkt_discard,
277 .output = ip6_pkt_discard_out,
279 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
280 .rt6i_protocol = RTPROT_KERNEL,
281 .rt6i_metric = ~(u32) 0,
282 .rt6i_ref = ATOMIC_INIT(1),
285 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
287 static const struct rt6_info ip6_prohibit_entry_template = {
289 .__refcnt = ATOMIC_INIT(1),
291 .obsolete = DST_OBSOLETE_FORCE_CHK,
293 .input = ip6_pkt_prohibit,
294 .output = ip6_pkt_prohibit_out,
296 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
297 .rt6i_protocol = RTPROT_KERNEL,
298 .rt6i_metric = ~(u32) 0,
299 .rt6i_ref = ATOMIC_INIT(1),
302 static const struct rt6_info ip6_blk_hole_entry_template = {
304 .__refcnt = ATOMIC_INIT(1),
306 .obsolete = DST_OBSOLETE_FORCE_CHK,
308 .input = dst_discard,
309 .output = dst_discard_out,
311 .rt6i_flags = (RTF_REJECT | RTF_NONEXTHOP),
312 .rt6i_protocol = RTPROT_KERNEL,
313 .rt6i_metric = ~(u32) 0,
314 .rt6i_ref = ATOMIC_INIT(1),
319 static void rt6_info_init(struct rt6_info *rt)
321 struct dst_entry *dst = &rt->dst;
323 memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
324 INIT_LIST_HEAD(&rt->rt6i_siblings);
325 INIT_LIST_HEAD(&rt->rt6i_uncached);
328 /* allocate dst with ip6_dst_ops */
329 static struct rt6_info *__ip6_dst_alloc(struct net *net,
330 struct net_device *dev,
333 struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
334 0, DST_OBSOLETE_FORCE_CHK, flags);
342 struct rt6_info *ip6_dst_alloc(struct net *net,
343 struct net_device *dev,
346 struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
349 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
353 for_each_possible_cpu(cpu) {
356 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
357 /* no one shares rt */
361 dst_destroy((struct dst_entry *)rt);
368 EXPORT_SYMBOL(ip6_dst_alloc);
370 static void ip6_dst_destroy(struct dst_entry *dst)
372 struct rt6_info *rt = (struct rt6_info *)dst;
373 struct dst_entry *from = dst->from;
374 struct inet6_dev *idev;
376 dst_destroy_metrics_generic(dst);
377 free_percpu(rt->rt6i_pcpu);
378 rt6_uncached_list_del(rt);
380 idev = rt->rt6i_idev;
382 rt->rt6i_idev = NULL;
390 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
393 struct rt6_info *rt = (struct rt6_info *)dst;
394 struct inet6_dev *idev = rt->rt6i_idev;
395 struct net_device *loopback_dev =
396 dev_net(dev)->loopback_dev;
398 if (dev != loopback_dev) {
399 if (idev && idev->dev == dev) {
400 struct inet6_dev *loopback_idev =
401 in6_dev_get(loopback_dev);
403 rt->rt6i_idev = loopback_idev;
410 static bool __rt6_check_expired(const struct rt6_info *rt)
412 if (rt->rt6i_flags & RTF_EXPIRES)
413 return time_after(jiffies, rt->dst.expires);
418 static bool rt6_check_expired(const struct rt6_info *rt)
420 if (rt->rt6i_flags & RTF_EXPIRES) {
421 if (time_after(jiffies, rt->dst.expires))
423 } else if (rt->dst.from) {
424 return rt6_check_expired((struct rt6_info *) rt->dst.from);
429 /* Multipath route selection:
430 * Hash based function using packet header and flowlabel.
431 * Adapted from fib_info_hashfn()
433 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
434 const struct flowi6 *fl6)
436 return get_hash_from_flowi6(fl6) % candidate_count;
439 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
440 struct flowi6 *fl6, int oif,
443 struct rt6_info *sibling, *next_sibling;
446 route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
447 /* Don't change the route, if route_choosen == 0
448 * (siblings does not include ourself)
451 list_for_each_entry_safe(sibling, next_sibling,
452 &match->rt6i_siblings, rt6i_siblings) {
454 if (route_choosen == 0) {
455 if (rt6_score_route(sibling, oif, strict) < 0)
465 * Route lookup. Any table->tb6_lock is implied.
468 static inline struct rt6_info *rt6_device_match(struct net *net,
470 const struct in6_addr *saddr,
474 struct rt6_info *local = NULL;
475 struct rt6_info *sprt;
477 if (!oif && ipv6_addr_any(saddr))
480 for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
481 struct net_device *dev = sprt->dst.dev;
484 if (dev->ifindex == oif)
486 if (dev->flags & IFF_LOOPBACK) {
487 if (!sprt->rt6i_idev ||
488 sprt->rt6i_idev->dev->ifindex != oif) {
489 if (flags & RT6_LOOKUP_F_IFACE)
492 local->rt6i_idev->dev->ifindex == oif)
498 if (ipv6_chk_addr(net, saddr, dev,
499 flags & RT6_LOOKUP_F_IFACE))
508 if (flags & RT6_LOOKUP_F_IFACE)
509 return net->ipv6.ip6_null_entry;
515 #ifdef CONFIG_IPV6_ROUTER_PREF
516 struct __rt6_probe_work {
517 struct work_struct work;
518 struct in6_addr target;
519 struct net_device *dev;
522 static void rt6_probe_deferred(struct work_struct *w)
524 struct in6_addr mcaddr;
525 struct __rt6_probe_work *work =
526 container_of(w, struct __rt6_probe_work, work);
528 addrconf_addr_solict_mult(&work->target, &mcaddr);
529 ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
534 static void rt6_probe(struct rt6_info *rt)
536 struct __rt6_probe_work *work;
537 struct neighbour *neigh;
539 * Okay, this does not seem to be appropriate
540 * for now, however, we need to check if it
541 * is really so; aka Router Reachability Probing.
543 * Router Reachability Probe MUST be rate-limited
544 * to no more than one per minute.
546 if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
549 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
551 if (neigh->nud_state & NUD_VALID)
555 write_lock(&neigh->lock);
556 if (!(neigh->nud_state & NUD_VALID) &&
559 rt->rt6i_idev->cnf.rtr_probe_interval)) {
560 work = kmalloc(sizeof(*work), GFP_ATOMIC);
562 __neigh_set_probe_once(neigh);
564 write_unlock(&neigh->lock);
566 work = kmalloc(sizeof(*work), GFP_ATOMIC);
570 INIT_WORK(&work->work, rt6_probe_deferred);
571 work->target = rt->rt6i_gateway;
572 dev_hold(rt->dst.dev);
573 work->dev = rt->dst.dev;
574 schedule_work(&work->work);
578 rcu_read_unlock_bh();
581 static inline void rt6_probe(struct rt6_info *rt)
587 * Default Router Selection (RFC 2461 6.3.6)
589 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
591 struct net_device *dev = rt->dst.dev;
592 if (!oif || dev->ifindex == oif)
594 if ((dev->flags & IFF_LOOPBACK) &&
595 rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
600 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
602 struct neighbour *neigh;
603 enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
605 if (rt->rt6i_flags & RTF_NONEXTHOP ||
606 !(rt->rt6i_flags & RTF_GATEWAY))
607 return RT6_NUD_SUCCEED;
610 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
612 read_lock(&neigh->lock);
613 if (neigh->nud_state & NUD_VALID)
614 ret = RT6_NUD_SUCCEED;
615 #ifdef CONFIG_IPV6_ROUTER_PREF
616 else if (!(neigh->nud_state & NUD_FAILED))
617 ret = RT6_NUD_SUCCEED;
619 ret = RT6_NUD_FAIL_PROBE;
621 read_unlock(&neigh->lock);
623 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
624 RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
626 rcu_read_unlock_bh();
631 static int rt6_score_route(struct rt6_info *rt, int oif,
636 m = rt6_check_dev(rt, oif);
637 if (!m && (strict & RT6_LOOKUP_F_IFACE))
638 return RT6_NUD_FAIL_HARD;
639 #ifdef CONFIG_IPV6_ROUTER_PREF
640 m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
642 if (strict & RT6_LOOKUP_F_REACHABLE) {
643 int n = rt6_check_neigh(rt);
650 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
651 int *mpri, struct rt6_info *match,
655 bool match_do_rr = false;
656 struct inet6_dev *idev = rt->rt6i_idev;
657 struct net_device *dev = rt->dst.dev;
659 if (dev && !netif_carrier_ok(dev) &&
660 idev->cnf.ignore_routes_with_linkdown)
663 if (rt6_check_expired(rt))
666 m = rt6_score_route(rt, oif, strict);
667 if (m == RT6_NUD_FAIL_DO_RR) {
669 m = 0; /* lowest valid score */
670 } else if (m == RT6_NUD_FAIL_HARD) {
674 if (strict & RT6_LOOKUP_F_REACHABLE)
677 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
679 *do_rr = match_do_rr;
687 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
688 struct rt6_info *rr_head,
689 u32 metric, int oif, int strict,
692 struct rt6_info *rt, *match, *cont;
697 for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
698 if (rt->rt6i_metric != metric) {
703 match = find_match(rt, oif, strict, &mpri, match, do_rr);
706 for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
707 if (rt->rt6i_metric != metric) {
712 match = find_match(rt, oif, strict, &mpri, match, do_rr);
718 for (rt = cont; rt; rt = rt->dst.rt6_next)
719 match = find_match(rt, oif, strict, &mpri, match, do_rr);
724 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
726 struct rt6_info *match, *rt0;
732 fn->rr_ptr = rt0 = fn->leaf;
734 match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
738 struct rt6_info *next = rt0->dst.rt6_next;
740 /* no entries matched; do round-robin */
741 if (!next || next->rt6i_metric != rt0->rt6i_metric)
748 net = dev_net(rt0->dst.dev);
749 return match ? match : net->ipv6.ip6_null_entry;
752 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
754 return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
757 #ifdef CONFIG_IPV6_ROUTE_INFO
758 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
759 const struct in6_addr *gwaddr)
761 struct net *net = dev_net(dev);
762 struct route_info *rinfo = (struct route_info *) opt;
763 struct in6_addr prefix_buf, *prefix;
765 unsigned long lifetime;
768 if (len < sizeof(struct route_info)) {
772 /* Sanity check for prefix_len and length */
773 if (rinfo->length > 3) {
775 } else if (rinfo->prefix_len > 128) {
777 } else if (rinfo->prefix_len > 64) {
778 if (rinfo->length < 2) {
781 } else if (rinfo->prefix_len > 0) {
782 if (rinfo->length < 1) {
787 pref = rinfo->route_pref;
788 if (pref == ICMPV6_ROUTER_PREF_INVALID)
791 lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
793 if (rinfo->length == 3)
794 prefix = (struct in6_addr *)rinfo->prefix;
796 /* this function is safe */
797 ipv6_addr_prefix(&prefix_buf,
798 (struct in6_addr *)rinfo->prefix,
800 prefix = &prefix_buf;
803 if (rinfo->prefix_len == 0)
804 rt = rt6_get_dflt_router(gwaddr, dev);
806 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
809 if (rt && !lifetime) {
815 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
818 rt->rt6i_flags = RTF_ROUTEINFO |
819 (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
822 if (!addrconf_finite_timeout(lifetime))
823 rt6_clean_expires(rt);
825 rt6_set_expires(rt, jiffies + HZ * lifetime);
833 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
834 struct in6_addr *saddr)
836 struct fib6_node *pn;
838 if (fn->fn_flags & RTN_TL_ROOT)
841 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
842 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
845 if (fn->fn_flags & RTN_RTINFO)
850 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
851 struct fib6_table *table,
852 struct flowi6 *fl6, int flags)
854 struct fib6_node *fn;
857 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
858 flags &= ~RT6_LOOKUP_F_IFACE;
860 read_lock_bh(&table->tb6_lock);
861 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
864 rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
865 if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
866 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
867 if (rt == net->ipv6.ip6_null_entry) {
868 fn = fib6_backtrack(fn, &fl6->saddr);
872 dst_use(&rt->dst, jiffies);
873 read_unlock_bh(&table->tb6_lock);
878 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
881 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
883 EXPORT_SYMBOL_GPL(ip6_route_lookup);
885 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
886 const struct in6_addr *saddr, int oif, int strict)
888 struct flowi6 fl6 = {
892 struct dst_entry *dst;
893 int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
896 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
897 flags |= RT6_LOOKUP_F_HAS_SADDR;
900 dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
902 return (struct rt6_info *) dst;
908 EXPORT_SYMBOL(rt6_lookup);
910 /* ip6_ins_rt is called with FREE table->tb6_lock.
911 It takes new route entry, the addition fails by any reason the
912 route is freed. In any case, if caller does not hold it, it may
916 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
917 struct mx6_config *mxc)
920 struct fib6_table *table;
922 table = rt->rt6i_table;
923 write_lock_bh(&table->tb6_lock);
924 err = fib6_add(&table->tb6_root, rt, info, mxc);
925 write_unlock_bh(&table->tb6_lock);
930 int ip6_ins_rt(struct rt6_info *rt)
932 struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
933 struct mx6_config mxc = { .mx = NULL, };
935 return __ip6_ins_rt(rt, &info, &mxc);
938 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
939 const struct in6_addr *daddr,
940 const struct in6_addr *saddr)
948 if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
949 ort = (struct rt6_info *)ort->dst.from;
951 rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
956 ip6_rt_copy_init(rt, ort);
957 rt->rt6i_flags |= RTF_CACHE;
959 rt->dst.flags |= DST_HOST;
960 rt->rt6i_dst.addr = *daddr;
961 rt->rt6i_dst.plen = 128;
963 if (!rt6_is_gw_or_nonexthop(ort)) {
964 if (ort->rt6i_dst.plen != 128 &&
965 ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
966 rt->rt6i_flags |= RTF_ANYCAST;
967 #ifdef CONFIG_IPV6_SUBTREES
968 if (rt->rt6i_src.plen && saddr) {
969 rt->rt6i_src.addr = *saddr;
970 rt->rt6i_src.plen = 128;
978 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
980 struct rt6_info *pcpu_rt;
982 pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
983 rt->dst.dev, rt->dst.flags);
987 ip6_rt_copy_init(pcpu_rt, rt);
988 pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
989 pcpu_rt->rt6i_flags |= RTF_PCPU;
993 /* It should be called with read_lock_bh(&tb6_lock) acquired */
994 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
996 struct rt6_info *pcpu_rt, **p;
998 p = this_cpu_ptr(rt->rt6i_pcpu);
1002 dst_hold(&pcpu_rt->dst);
1003 rt6_dst_from_metrics_check(pcpu_rt);
1008 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1010 struct fib6_table *table = rt->rt6i_table;
1011 struct rt6_info *pcpu_rt, *prev, **p;
1013 pcpu_rt = ip6_rt_pcpu_alloc(rt);
1015 struct net *net = dev_net(rt->dst.dev);
1017 dst_hold(&net->ipv6.ip6_null_entry->dst);
1018 return net->ipv6.ip6_null_entry;
1021 read_lock_bh(&table->tb6_lock);
1022 if (rt->rt6i_pcpu) {
1023 p = this_cpu_ptr(rt->rt6i_pcpu);
1024 prev = cmpxchg(p, NULL, pcpu_rt);
1026 /* If someone did it before us, return prev instead */
1027 dst_destroy(&pcpu_rt->dst);
1031 /* rt has been removed from the fib6 tree
1032 * before we have a chance to acquire the read_lock.
1033 * In this case, don't brother to create a pcpu rt
1034 * since rt is going away anyway. The next
1035 * dst_check() will trigger a re-lookup.
1037 dst_destroy(&pcpu_rt->dst);
1040 dst_hold(&pcpu_rt->dst);
1041 rt6_dst_from_metrics_check(pcpu_rt);
1042 read_unlock_bh(&table->tb6_lock);
1046 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1047 struct flowi6 *fl6, int flags)
1049 struct fib6_node *fn, *saved_fn;
1050 struct rt6_info *rt;
1053 strict |= flags & RT6_LOOKUP_F_IFACE;
1054 if (net->ipv6.devconf_all->forwarding == 0)
1055 strict |= RT6_LOOKUP_F_REACHABLE;
1057 read_lock_bh(&table->tb6_lock);
1059 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1062 if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1066 rt = rt6_select(fn, oif, strict);
1067 if (rt->rt6i_nsiblings)
1068 rt = rt6_multipath_select(rt, fl6, oif, strict);
1069 if (rt == net->ipv6.ip6_null_entry) {
1070 fn = fib6_backtrack(fn, &fl6->saddr);
1072 goto redo_rt6_select;
1073 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1074 /* also consider unreachable route */
1075 strict &= ~RT6_LOOKUP_F_REACHABLE;
1077 goto redo_rt6_select;
1082 if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1083 dst_use(&rt->dst, jiffies);
1084 read_unlock_bh(&table->tb6_lock);
1086 rt6_dst_from_metrics_check(rt);
1088 } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1089 !(rt->rt6i_flags & RTF_GATEWAY))) {
1090 /* Create a RTF_CACHE clone which will not be
1091 * owned by the fib6 tree. It is for the special case where
1092 * the daddr in the skb during the neighbor look-up is different
1093 * from the fl6->daddr used to look-up route here.
1096 struct rt6_info *uncached_rt;
1098 dst_use(&rt->dst, jiffies);
1099 read_unlock_bh(&table->tb6_lock);
1101 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1102 dst_release(&rt->dst);
1105 rt6_uncached_list_add(uncached_rt);
1107 uncached_rt = net->ipv6.ip6_null_entry;
1109 dst_hold(&uncached_rt->dst);
1113 /* Get a percpu copy */
1115 struct rt6_info *pcpu_rt;
1117 rt->dst.lastuse = jiffies;
1119 pcpu_rt = rt6_get_pcpu_route(rt);
1122 read_unlock_bh(&table->tb6_lock);
1124 /* We have to do the read_unlock first
1125 * because rt6_make_pcpu_route() may trigger
1126 * ip6_dst_gc() which will take the write_lock.
1129 read_unlock_bh(&table->tb6_lock);
1130 pcpu_rt = rt6_make_pcpu_route(rt);
1131 dst_release(&rt->dst);
1139 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1140 struct flowi6 *fl6, int flags)
1142 return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1145 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1146 struct net_device *dev,
1147 struct flowi6 *fl6, int flags)
1149 if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1150 flags |= RT6_LOOKUP_F_IFACE;
1152 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1155 void ip6_route_input(struct sk_buff *skb)
1157 const struct ipv6hdr *iph = ipv6_hdr(skb);
1158 struct net *net = dev_net(skb->dev);
1159 int flags = RT6_LOOKUP_F_HAS_SADDR;
1160 struct ip_tunnel_info *tun_info;
1161 struct flowi6 fl6 = {
1162 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1163 .daddr = iph->daddr,
1164 .saddr = iph->saddr,
1165 .flowlabel = ip6_flowinfo(iph),
1166 .flowi6_mark = skb->mark,
1167 .flowi6_proto = iph->nexthdr,
1170 tun_info = skb_tunnel_info(skb);
1171 if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1172 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1174 skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1177 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1178 struct flowi6 *fl6, int flags)
1180 return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1183 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1184 struct flowi6 *fl6, int flags)
1186 struct dst_entry *dst;
1189 dst = l3mdev_rt6_dst_by_oif(net, fl6);
1193 fl6->flowi6_iif = LOOPBACK_IFINDEX;
1195 any_src = ipv6_addr_any(&fl6->saddr);
1196 if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1197 (fl6->flowi6_oif && any_src))
1198 flags |= RT6_LOOKUP_F_IFACE;
1201 flags |= RT6_LOOKUP_F_HAS_SADDR;
1203 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1205 return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1207 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1209 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1211 struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1212 struct dst_entry *new = NULL;
1214 rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1220 new->input = dst_discard;
1221 new->output = dst_discard_out;
1223 dst_copy_metrics(new, &ort->dst);
1224 rt->rt6i_idev = ort->rt6i_idev;
1226 in6_dev_hold(rt->rt6i_idev);
1228 rt->rt6i_gateway = ort->rt6i_gateway;
1229 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1230 rt->rt6i_metric = 0;
1232 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1233 #ifdef CONFIG_IPV6_SUBTREES
1234 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1240 dst_release(dst_orig);
1241 return new ? new : ERR_PTR(-ENOMEM);
1245 * Destination cache support functions
1248 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1251 dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1252 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1255 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1259 if (!rt6_get_cookie_safe(rt, &rt_cookie) || rt_cookie != cookie)
1262 if (rt6_check_expired(rt))
1268 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1270 if (!__rt6_check_expired(rt) &&
1271 rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1272 rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1278 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1280 struct rt6_info *rt;
1282 rt = (struct rt6_info *) dst;
1284 /* All IPV6 dsts are created with ->obsolete set to the value
1285 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1286 * into this function always.
1289 rt6_dst_from_metrics_check(rt);
1291 if (rt->rt6i_flags & RTF_PCPU ||
1292 (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1293 return rt6_dst_from_check(rt, cookie);
1295 return rt6_check(rt, cookie);
1298 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1300 struct rt6_info *rt = (struct rt6_info *) dst;
1303 if (rt->rt6i_flags & RTF_CACHE) {
1304 if (rt6_check_expired(rt)) {
1316 static void ip6_link_failure(struct sk_buff *skb)
1318 struct rt6_info *rt;
1320 icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1322 rt = (struct rt6_info *) skb_dst(skb);
1324 if (rt->rt6i_flags & RTF_CACHE) {
1328 struct fib6_node *fn;
1331 fn = rcu_dereference(rt->rt6i_node);
1332 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
1339 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1341 struct net *net = dev_net(rt->dst.dev);
1343 rt->rt6i_flags |= RTF_MODIFIED;
1344 rt->rt6i_pmtu = mtu;
1345 rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1348 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1350 return !(rt->rt6i_flags & RTF_CACHE) &&
1351 (rt->rt6i_flags & RTF_PCPU ||
1352 rcu_access_pointer(rt->rt6i_node));
1355 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1356 const struct ipv6hdr *iph, u32 mtu)
1358 struct rt6_info *rt6 = (struct rt6_info *)dst;
1360 if (rt6->rt6i_flags & RTF_LOCAL)
1364 mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1365 if (mtu >= dst_mtu(dst))
1368 if (!rt6_cache_allowed_for_pmtu(rt6)) {
1369 rt6_do_update_pmtu(rt6, mtu);
1371 const struct in6_addr *daddr, *saddr;
1372 struct rt6_info *nrt6;
1375 daddr = &iph->daddr;
1376 saddr = &iph->saddr;
1378 daddr = &sk->sk_v6_daddr;
1379 saddr = &inet6_sk(sk)->saddr;
1383 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1385 rt6_do_update_pmtu(nrt6, mtu);
1387 /* ip6_ins_rt(nrt6) will bump the
1388 * rt6->rt6i_node->fn_sernum
1389 * which will fail the next rt6_check() and
1390 * invalidate the sk->sk_dst_cache.
1397 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1398 struct sk_buff *skb, u32 mtu)
1400 __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1403 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1406 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1407 struct dst_entry *dst;
1410 memset(&fl6, 0, sizeof(fl6));
1411 fl6.flowi6_oif = oif;
1412 fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1413 fl6.daddr = iph->daddr;
1414 fl6.saddr = iph->saddr;
1415 fl6.flowlabel = ip6_flowinfo(iph);
1417 dst = ip6_route_output(net, NULL, &fl6);
1419 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1422 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1424 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1426 int oif = sk->sk_bound_dev_if;
1428 if (!oif && skb->dev)
1429 oif = l3mdev_master_ifindex(skb->dev);
1431 ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark);
1433 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1435 /* Handle redirects */
1436 struct ip6rd_flowi {
1438 struct in6_addr gateway;
1441 static struct rt6_info *__ip6_route_redirect(struct net *net,
1442 struct fib6_table *table,
1446 struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1447 struct rt6_info *rt;
1448 struct fib6_node *fn;
1450 /* Get the "current" route for this destination and
1451 * check if the redirect has come from approriate router.
1453 * RFC 4861 specifies that redirects should only be
1454 * accepted if they come from the nexthop to the target.
1455 * Due to the way the routes are chosen, this notion
1456 * is a bit fuzzy and one might need to check all possible
1460 read_lock_bh(&table->tb6_lock);
1461 fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1463 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1464 if (rt6_check_expired(rt))
1468 if (!(rt->rt6i_flags & RTF_GATEWAY))
1470 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1472 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1478 rt = net->ipv6.ip6_null_entry;
1479 else if (rt->dst.error) {
1480 rt = net->ipv6.ip6_null_entry;
1484 if (rt == net->ipv6.ip6_null_entry) {
1485 fn = fib6_backtrack(fn, &fl6->saddr);
1493 read_unlock_bh(&table->tb6_lock);
1498 static struct dst_entry *ip6_route_redirect(struct net *net,
1499 const struct flowi6 *fl6,
1500 const struct in6_addr *gateway)
1502 int flags = RT6_LOOKUP_F_HAS_SADDR;
1503 struct ip6rd_flowi rdfl;
1506 rdfl.gateway = *gateway;
1508 return fib6_rule_lookup(net, &rdfl.fl6,
1509 flags, __ip6_route_redirect);
1512 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1514 const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1515 struct dst_entry *dst;
1518 memset(&fl6, 0, sizeof(fl6));
1519 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1520 fl6.flowi6_oif = oif;
1521 fl6.flowi6_mark = mark;
1522 fl6.daddr = iph->daddr;
1523 fl6.saddr = iph->saddr;
1524 fl6.flowlabel = ip6_flowinfo(iph);
1526 dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1527 rt6_do_redirect(dst, NULL, skb);
1530 EXPORT_SYMBOL_GPL(ip6_redirect);
1532 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1535 const struct ipv6hdr *iph = ipv6_hdr(skb);
1536 const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1537 struct dst_entry *dst;
1540 memset(&fl6, 0, sizeof(fl6));
1541 fl6.flowi6_iif = LOOPBACK_IFINDEX;
1542 fl6.flowi6_oif = oif;
1543 fl6.flowi6_mark = mark;
1544 fl6.daddr = msg->dest;
1545 fl6.saddr = iph->daddr;
1547 dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1548 rt6_do_redirect(dst, NULL, skb);
1552 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1554 ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1556 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1558 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1560 struct net_device *dev = dst->dev;
1561 unsigned int mtu = dst_mtu(dst);
1562 struct net *net = dev_net(dev);
1564 mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1566 if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1567 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1570 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1571 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1572 * IPV6_MAXPLEN is also valid and means: "any MSS,
1573 * rely only on pmtu discovery"
1575 if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1580 static unsigned int ip6_mtu(const struct dst_entry *dst)
1582 const struct rt6_info *rt = (const struct rt6_info *)dst;
1583 unsigned int mtu = rt->rt6i_pmtu;
1584 struct inet6_dev *idev;
1589 mtu = dst_metric_raw(dst, RTAX_MTU);
1596 idev = __in6_dev_get(dst->dev);
1598 mtu = idev->cnf.mtu6;
1602 return min_t(unsigned int, mtu, IP6_MAX_MTU);
1605 static struct dst_entry *icmp6_dst_gc_list;
1606 static DEFINE_SPINLOCK(icmp6_dst_lock);
1608 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1611 struct dst_entry *dst;
1612 struct rt6_info *rt;
1613 struct inet6_dev *idev = in6_dev_get(dev);
1614 struct net *net = dev_net(dev);
1616 if (unlikely(!idev))
1617 return ERR_PTR(-ENODEV);
1619 rt = ip6_dst_alloc(net, dev, 0);
1620 if (unlikely(!rt)) {
1622 dst = ERR_PTR(-ENOMEM);
1626 rt->dst.flags |= DST_HOST;
1627 rt->dst.input = ip6_input;
1628 rt->dst.output = ip6_output;
1629 atomic_set(&rt->dst.__refcnt, 1);
1630 rt->rt6i_gateway = fl6->daddr;
1631 rt->rt6i_dst.addr = fl6->daddr;
1632 rt->rt6i_dst.plen = 128;
1633 rt->rt6i_idev = idev;
1634 dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1636 spin_lock_bh(&icmp6_dst_lock);
1637 rt->dst.next = icmp6_dst_gc_list;
1638 icmp6_dst_gc_list = &rt->dst;
1639 spin_unlock_bh(&icmp6_dst_lock);
1641 fib6_force_start_gc(net);
1643 dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1649 int icmp6_dst_gc(void)
1651 struct dst_entry *dst, **pprev;
1654 spin_lock_bh(&icmp6_dst_lock);
1655 pprev = &icmp6_dst_gc_list;
1657 while ((dst = *pprev) != NULL) {
1658 if (!atomic_read(&dst->__refcnt)) {
1667 spin_unlock_bh(&icmp6_dst_lock);
1672 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1675 struct dst_entry *dst, **pprev;
1677 spin_lock_bh(&icmp6_dst_lock);
1678 pprev = &icmp6_dst_gc_list;
1679 while ((dst = *pprev) != NULL) {
1680 struct rt6_info *rt = (struct rt6_info *) dst;
1681 if (func(rt, arg)) {
1688 spin_unlock_bh(&icmp6_dst_lock);
1691 static int ip6_dst_gc(struct dst_ops *ops)
1693 struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1694 int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1695 int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1696 int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1697 int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1698 unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1701 entries = dst_entries_get_fast(ops);
1702 if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1703 entries <= rt_max_size)
1706 net->ipv6.ip6_rt_gc_expire++;
1707 fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1708 entries = dst_entries_get_slow(ops);
1709 if (entries < ops->gc_thresh)
1710 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1712 net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1713 return entries > rt_max_size;
1716 static int ip6_convert_metrics(struct mx6_config *mxc,
1717 const struct fib6_config *cfg)
1719 bool ecn_ca = false;
1727 mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1731 nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1732 int type = nla_type(nla);
1737 if (unlikely(type > RTAX_MAX))
1740 if (type == RTAX_CC_ALGO) {
1741 char tmp[TCP_CA_NAME_MAX];
1743 nla_strlcpy(tmp, nla, sizeof(tmp));
1744 val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1745 if (val == TCP_CA_UNSPEC)
1748 val = nla_get_u32(nla);
1750 if (type == RTAX_HOPLIMIT && val > 255)
1752 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1756 __set_bit(type - 1, mxc->mx_valid);
1760 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1761 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1771 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1773 struct net *net = cfg->fc_nlinfo.nl_net;
1774 struct rt6_info *rt = NULL;
1775 struct net_device *dev = NULL;
1776 struct inet6_dev *idev = NULL;
1777 struct fib6_table *table;
1781 /* RTF_PCPU is an internal flag; can not be set by userspace */
1782 if (cfg->fc_flags & RTF_PCPU)
1785 if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1787 #ifndef CONFIG_IPV6_SUBTREES
1788 if (cfg->fc_src_len)
1791 if (cfg->fc_ifindex) {
1793 dev = dev_get_by_index(net, cfg->fc_ifindex);
1796 idev = in6_dev_get(dev);
1801 if (cfg->fc_metric == 0)
1802 cfg->fc_metric = IP6_RT_PRIO_USER;
1805 if (cfg->fc_nlinfo.nlh &&
1806 !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1807 table = fib6_get_table(net, cfg->fc_table);
1809 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1810 table = fib6_new_table(net, cfg->fc_table);
1813 table = fib6_new_table(net, cfg->fc_table);
1819 rt = ip6_dst_alloc(net, NULL,
1820 (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1827 if (cfg->fc_flags & RTF_EXPIRES)
1828 rt6_set_expires(rt, jiffies +
1829 clock_t_to_jiffies(cfg->fc_expires));
1831 rt6_clean_expires(rt);
1833 if (cfg->fc_protocol == RTPROT_UNSPEC)
1834 cfg->fc_protocol = RTPROT_BOOT;
1835 rt->rt6i_protocol = cfg->fc_protocol;
1837 addr_type = ipv6_addr_type(&cfg->fc_dst);
1839 if (addr_type & IPV6_ADDR_MULTICAST)
1840 rt->dst.input = ip6_mc_input;
1841 else if (cfg->fc_flags & RTF_LOCAL)
1842 rt->dst.input = ip6_input;
1844 rt->dst.input = ip6_forward;
1846 rt->dst.output = ip6_output;
1848 if (cfg->fc_encap) {
1849 struct lwtunnel_state *lwtstate;
1851 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1852 cfg->fc_encap, AF_INET6, cfg,
1856 rt->dst.lwtstate = lwtstate_get(lwtstate);
1857 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1858 rt->dst.lwtstate->orig_output = rt->dst.output;
1859 rt->dst.output = lwtunnel_output;
1861 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1862 rt->dst.lwtstate->orig_input = rt->dst.input;
1863 rt->dst.input = lwtunnel_input;
1867 ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1868 rt->rt6i_dst.plen = cfg->fc_dst_len;
1869 if (rt->rt6i_dst.plen == 128)
1870 rt->dst.flags |= DST_HOST;
1872 #ifdef CONFIG_IPV6_SUBTREES
1873 ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1874 rt->rt6i_src.plen = cfg->fc_src_len;
1877 rt->rt6i_metric = cfg->fc_metric;
1879 /* We cannot add true routes via loopback here,
1880 they would result in kernel looping; promote them to reject routes
1882 if ((cfg->fc_flags & RTF_REJECT) ||
1883 (dev && (dev->flags & IFF_LOOPBACK) &&
1884 !(addr_type & IPV6_ADDR_LOOPBACK) &&
1885 !(cfg->fc_flags & RTF_LOCAL))) {
1886 /* hold loopback dev/idev if we haven't done so. */
1887 if (dev != net->loopback_dev) {
1892 dev = net->loopback_dev;
1894 idev = in6_dev_get(dev);
1900 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1901 switch (cfg->fc_type) {
1903 rt->dst.error = -EINVAL;
1904 rt->dst.output = dst_discard_out;
1905 rt->dst.input = dst_discard;
1908 rt->dst.error = -EACCES;
1909 rt->dst.output = ip6_pkt_prohibit_out;
1910 rt->dst.input = ip6_pkt_prohibit;
1913 case RTN_UNREACHABLE:
1915 rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1916 : (cfg->fc_type == RTN_UNREACHABLE)
1917 ? -EHOSTUNREACH : -ENETUNREACH;
1918 rt->dst.output = ip6_pkt_discard_out;
1919 rt->dst.input = ip6_pkt_discard;
1925 if (cfg->fc_flags & RTF_GATEWAY) {
1926 const struct in6_addr *gw_addr;
1929 gw_addr = &cfg->fc_gateway;
1930 gwa_type = ipv6_addr_type(gw_addr);
1932 /* if gw_addr is local we will fail to detect this in case
1933 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1934 * will return already-added prefix route via interface that
1935 * prefix route was assigned to, which might be non-loopback.
1938 if (ipv6_chk_addr_and_flags(net, gw_addr,
1939 gwa_type & IPV6_ADDR_LINKLOCAL ?
1943 rt->rt6i_gateway = *gw_addr;
1945 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1946 struct rt6_info *grt;
1948 /* IPv6 strictly inhibits using not link-local
1949 addresses as nexthop address.
1950 Otherwise, router will not able to send redirects.
1951 It is very good, but in some (rare!) circumstances
1952 (SIT, PtP, NBMA NOARP links) it is handy to allow
1953 some exceptions. --ANK
1955 if (!(gwa_type & IPV6_ADDR_UNICAST))
1958 grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1960 err = -EHOSTUNREACH;
1964 if (dev != grt->dst.dev) {
1970 idev = grt->rt6i_idev;
1972 in6_dev_hold(grt->rt6i_idev);
1974 if (!(grt->rt6i_flags & RTF_GATEWAY))
1982 if (!dev || (dev->flags & IFF_LOOPBACK))
1990 if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1991 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1995 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1996 rt->rt6i_prefsrc.plen = 128;
1998 rt->rt6i_prefsrc.plen = 0;
2000 rt->rt6i_flags = cfg->fc_flags;
2004 rt->rt6i_idev = idev;
2005 rt->rt6i_table = table;
2007 cfg->fc_nlinfo.nl_net = dev_net(dev);
2018 return ERR_PTR(err);
2021 int ip6_route_add(struct fib6_config *cfg)
2023 struct mx6_config mxc = { .mx = NULL, };
2024 struct rt6_info *rt;
2027 rt = ip6_route_info_create(cfg);
2034 err = ip6_convert_metrics(&mxc, cfg);
2038 err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2050 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2053 struct fib6_table *table;
2054 struct net *net = dev_net(rt->dst.dev);
2056 if (rt == net->ipv6.ip6_null_entry ||
2057 rt->dst.flags & DST_NOCACHE) {
2062 table = rt->rt6i_table;
2063 write_lock_bh(&table->tb6_lock);
2064 err = fib6_del(rt, info);
2065 write_unlock_bh(&table->tb6_lock);
2072 int ip6_del_rt(struct rt6_info *rt)
2074 struct nl_info info = {
2075 .nl_net = dev_net(rt->dst.dev),
2077 return __ip6_del_rt(rt, &info);
2080 static int ip6_route_del(struct fib6_config *cfg)
2082 struct fib6_table *table;
2083 struct fib6_node *fn;
2084 struct rt6_info *rt;
2087 table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2091 read_lock_bh(&table->tb6_lock);
2093 fn = fib6_locate(&table->tb6_root,
2094 &cfg->fc_dst, cfg->fc_dst_len,
2095 &cfg->fc_src, cfg->fc_src_len);
2098 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2099 if ((rt->rt6i_flags & RTF_CACHE) &&
2100 !(cfg->fc_flags & RTF_CACHE))
2102 if (cfg->fc_ifindex &&
2104 rt->dst.dev->ifindex != cfg->fc_ifindex))
2106 if (cfg->fc_flags & RTF_GATEWAY &&
2107 !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2109 if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2111 if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2114 read_unlock_bh(&table->tb6_lock);
2116 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2119 read_unlock_bh(&table->tb6_lock);
2124 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2126 struct netevent_redirect netevent;
2127 struct rt6_info *rt, *nrt = NULL;
2128 struct ndisc_options ndopts;
2129 struct inet6_dev *in6_dev;
2130 struct neighbour *neigh;
2132 int optlen, on_link;
2135 optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2136 optlen -= sizeof(*msg);
2139 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2143 msg = (struct rd_msg *)icmp6_hdr(skb);
2145 if (ipv6_addr_is_multicast(&msg->dest)) {
2146 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2151 if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2153 } else if (ipv6_addr_type(&msg->target) !=
2154 (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2155 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2159 in6_dev = __in6_dev_get(skb->dev);
2162 if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2166 * The IP source address of the Redirect MUST be the same as the current
2167 * first-hop router for the specified ICMP Destination Address.
2170 if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2171 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2176 if (ndopts.nd_opts_tgt_lladdr) {
2177 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2180 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2185 rt = (struct rt6_info *) dst;
2186 if (rt->rt6i_flags & RTF_REJECT) {
2187 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2191 /* Redirect received -> path was valid.
2192 * Look, redirects are sent only in response to data packets,
2193 * so that this nexthop apparently is reachable. --ANK
2195 dst_confirm(&rt->dst);
2197 neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2202 * We have finally decided to accept it.
2205 neigh_update(neigh, lladdr, NUD_STALE,
2206 NEIGH_UPDATE_F_WEAK_OVERRIDE|
2207 NEIGH_UPDATE_F_OVERRIDE|
2208 (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2209 NEIGH_UPDATE_F_ISROUTER))
2212 nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2216 nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2218 nrt->rt6i_flags &= ~RTF_GATEWAY;
2220 nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2222 if (ip6_ins_rt(nrt))
2225 netevent.old = &rt->dst;
2226 netevent.new = &nrt->dst;
2227 netevent.daddr = &msg->dest;
2228 netevent.neigh = neigh;
2229 call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2231 if (rt->rt6i_flags & RTF_CACHE) {
2232 rt = (struct rt6_info *) dst_clone(&rt->dst);
2237 neigh_release(neigh);
2241 * Misc support functions
2244 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2246 BUG_ON(from->dst.from);
2248 rt->rt6i_flags &= ~RTF_EXPIRES;
2249 dst_hold(&from->dst);
2250 rt->dst.from = &from->dst;
2251 dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2254 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2256 rt->dst.input = ort->dst.input;
2257 rt->dst.output = ort->dst.output;
2258 rt->rt6i_dst = ort->rt6i_dst;
2259 rt->dst.error = ort->dst.error;
2260 rt->rt6i_idev = ort->rt6i_idev;
2262 in6_dev_hold(rt->rt6i_idev);
2263 rt->dst.lastuse = jiffies;
2264 rt->rt6i_gateway = ort->rt6i_gateway;
2265 rt->rt6i_flags = ort->rt6i_flags;
2266 rt6_set_from(rt, ort);
2267 rt->rt6i_metric = ort->rt6i_metric;
2268 #ifdef CONFIG_IPV6_SUBTREES
2269 rt->rt6i_src = ort->rt6i_src;
2271 rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2272 rt->rt6i_table = ort->rt6i_table;
2273 rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2276 #ifdef CONFIG_IPV6_ROUTE_INFO
2277 static struct rt6_info *rt6_get_route_info(struct net *net,
2278 const struct in6_addr *prefix, int prefixlen,
2279 const struct in6_addr *gwaddr,
2280 struct net_device *dev)
2282 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2283 int ifindex = dev->ifindex;
2284 struct fib6_node *fn;
2285 struct rt6_info *rt = NULL;
2286 struct fib6_table *table;
2288 table = fib6_get_table(net, tb_id);
2292 read_lock_bh(&table->tb6_lock);
2293 fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2297 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2298 if (rt->dst.dev->ifindex != ifindex)
2300 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2302 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2308 read_unlock_bh(&table->tb6_lock);
2312 static struct rt6_info *rt6_add_route_info(struct net *net,
2313 const struct in6_addr *prefix, int prefixlen,
2314 const struct in6_addr *gwaddr,
2315 struct net_device *dev,
2318 struct fib6_config cfg = {
2319 .fc_metric = IP6_RT_PRIO_USER,
2320 .fc_ifindex = dev->ifindex,
2321 .fc_dst_len = prefixlen,
2322 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2323 RTF_UP | RTF_PREF(pref),
2324 .fc_nlinfo.portid = 0,
2325 .fc_nlinfo.nlh = NULL,
2326 .fc_nlinfo.nl_net = net,
2329 cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2330 cfg.fc_dst = *prefix;
2331 cfg.fc_gateway = *gwaddr;
2333 /* We should treat it as a default route if prefix length is 0. */
2335 cfg.fc_flags |= RTF_DEFAULT;
2337 ip6_route_add(&cfg);
2339 return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2343 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2345 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2346 struct rt6_info *rt;
2347 struct fib6_table *table;
2349 table = fib6_get_table(dev_net(dev), tb_id);
2353 read_lock_bh(&table->tb6_lock);
2354 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2355 if (dev == rt->dst.dev &&
2356 ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2357 ipv6_addr_equal(&rt->rt6i_gateway, addr))
2362 read_unlock_bh(&table->tb6_lock);
2366 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2367 struct net_device *dev,
2370 struct fib6_config cfg = {
2371 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2372 .fc_metric = IP6_RT_PRIO_USER,
2373 .fc_ifindex = dev->ifindex,
2374 .fc_flags = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2375 RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2376 .fc_nlinfo.portid = 0,
2377 .fc_nlinfo.nlh = NULL,
2378 .fc_nlinfo.nl_net = dev_net(dev),
2381 cfg.fc_gateway = *gwaddr;
2383 if (!ip6_route_add(&cfg)) {
2384 struct fib6_table *table;
2386 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2388 table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2391 return rt6_get_dflt_router(gwaddr, dev);
2394 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2396 struct rt6_info *rt;
2399 read_lock_bh(&table->tb6_lock);
2400 for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2401 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2402 (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2404 read_unlock_bh(&table->tb6_lock);
2409 read_unlock_bh(&table->tb6_lock);
2411 table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2414 void rt6_purge_dflt_routers(struct net *net)
2416 struct fib6_table *table;
2417 struct hlist_head *head;
2422 for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2423 head = &net->ipv6.fib_table_hash[h];
2424 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2425 if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2426 __rt6_purge_dflt_routers(table);
2433 static void rtmsg_to_fib6_config(struct net *net,
2434 struct in6_rtmsg *rtmsg,
2435 struct fib6_config *cfg)
2437 memset(cfg, 0, sizeof(*cfg));
2439 cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2441 cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2442 cfg->fc_metric = rtmsg->rtmsg_metric;
2443 cfg->fc_expires = rtmsg->rtmsg_info;
2444 cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2445 cfg->fc_src_len = rtmsg->rtmsg_src_len;
2446 cfg->fc_flags = rtmsg->rtmsg_flags;
2448 cfg->fc_nlinfo.nl_net = net;
2450 cfg->fc_dst = rtmsg->rtmsg_dst;
2451 cfg->fc_src = rtmsg->rtmsg_src;
2452 cfg->fc_gateway = rtmsg->rtmsg_gateway;
2455 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2457 struct fib6_config cfg;
2458 struct in6_rtmsg rtmsg;
2462 case SIOCADDRT: /* Add a route */
2463 case SIOCDELRT: /* Delete a route */
2464 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2466 err = copy_from_user(&rtmsg, arg,
2467 sizeof(struct in6_rtmsg));
2471 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2476 err = ip6_route_add(&cfg);
2479 err = ip6_route_del(&cfg);
2493 * Drop the packet on the floor
2496 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2499 struct dst_entry *dst = skb_dst(skb);
2500 switch (ipstats_mib_noroutes) {
2501 case IPSTATS_MIB_INNOROUTES:
2502 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2503 if (type == IPV6_ADDR_ANY) {
2504 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2505 IPSTATS_MIB_INADDRERRORS);
2509 case IPSTATS_MIB_OUTNOROUTES:
2510 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2511 ipstats_mib_noroutes);
2514 icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2519 static int ip6_pkt_discard(struct sk_buff *skb)
2521 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2524 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2526 skb->dev = skb_dst(skb)->dev;
2527 return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2530 static int ip6_pkt_prohibit(struct sk_buff *skb)
2532 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2535 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2537 skb->dev = skb_dst(skb)->dev;
2538 return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2542 * Allocate a dst for local (unicast / anycast) address.
2545 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2546 const struct in6_addr *addr,
2550 struct net *net = dev_net(idev->dev);
2551 struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2554 return ERR_PTR(-ENOMEM);
2558 rt->dst.flags |= DST_HOST;
2559 rt->dst.input = ip6_input;
2560 rt->dst.output = ip6_output;
2561 rt->rt6i_idev = idev;
2563 rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2565 rt->rt6i_flags |= RTF_ANYCAST;
2567 rt->rt6i_flags |= RTF_LOCAL;
2569 rt->rt6i_gateway = *addr;
2570 rt->rt6i_dst.addr = *addr;
2571 rt->rt6i_dst.plen = 128;
2572 tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2573 rt->rt6i_table = fib6_get_table(net, tb_id);
2574 rt->dst.flags |= DST_NOCACHE;
2576 atomic_set(&rt->dst.__refcnt, 1);
2581 int ip6_route_get_saddr(struct net *net,
2582 struct rt6_info *rt,
2583 const struct in6_addr *daddr,
2585 struct in6_addr *saddr)
2587 struct inet6_dev *idev =
2588 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2590 if (rt && rt->rt6i_prefsrc.plen)
2591 *saddr = rt->rt6i_prefsrc.addr;
2593 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2594 daddr, prefs, saddr);
2598 /* remove deleted ip from prefsrc entries */
2599 struct arg_dev_net_ip {
2600 struct net_device *dev;
2602 struct in6_addr *addr;
2605 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2607 struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2608 struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2609 struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2611 if (((void *)rt->dst.dev == dev || !dev) &&
2612 rt != net->ipv6.ip6_null_entry &&
2613 ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2614 /* remove prefsrc entry */
2615 rt->rt6i_prefsrc.plen = 0;
2620 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2622 struct net *net = dev_net(ifp->idev->dev);
2623 struct arg_dev_net_ip adni = {
2624 .dev = ifp->idev->dev,
2628 fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2631 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2632 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2634 /* Remove routers and update dst entries when gateway turn into host. */
2635 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2637 struct in6_addr *gateway = (struct in6_addr *)arg;
2639 if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2640 ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2641 ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2647 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2649 fib6_clean_all(net, fib6_clean_tohost, gateway);
2652 struct arg_dev_net {
2653 struct net_device *dev;
2657 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2659 const struct arg_dev_net *adn = arg;
2660 const struct net_device *dev = adn->dev;
2662 if ((rt->dst.dev == dev || !dev) &&
2663 rt != adn->net->ipv6.ip6_null_entry)
2669 void rt6_ifdown(struct net *net, struct net_device *dev)
2671 struct arg_dev_net adn = {
2676 fib6_clean_all(net, fib6_ifdown, &adn);
2677 icmp6_clean_all(fib6_ifdown, &adn);
2679 rt6_uncached_list_flush_dev(net, dev);
2682 struct rt6_mtu_change_arg {
2683 struct net_device *dev;
2687 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2689 struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2690 struct inet6_dev *idev;
2692 /* In IPv6 pmtu discovery is not optional,
2693 so that RTAX_MTU lock cannot disable it.
2694 We still use this lock to block changes
2695 caused by addrconf/ndisc.
2698 idev = __in6_dev_get(arg->dev);
2702 /* For administrative MTU increase, there is no way to discover
2703 IPv6 PMTU increase, so PMTU increase should be updated here.
2704 Since RFC 1981 doesn't include administrative MTU increase
2705 update PMTU increase is a MUST. (i.e. jumbo frame)
2708 If new MTU is less than route PMTU, this new MTU will be the
2709 lowest MTU in the path, update the route PMTU to reflect PMTU
2710 decreases; if new MTU is greater than route PMTU, and the
2711 old MTU is the lowest MTU in the path, update the route PMTU
2712 to reflect the increase. In this case if the other nodes' MTU
2713 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2716 if (rt->dst.dev == arg->dev &&
2717 !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2718 if (rt->rt6i_flags & RTF_CACHE) {
2719 /* For RTF_CACHE with rt6i_pmtu == 0
2720 * (i.e. a redirected route),
2721 * the metrics of its rt->dst.from has already
2724 if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2725 rt->rt6i_pmtu = arg->mtu;
2726 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2727 (dst_mtu(&rt->dst) < arg->mtu &&
2728 dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2729 dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2735 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2737 struct rt6_mtu_change_arg arg = {
2742 fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2745 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2746 [RTA_GATEWAY] = { .len = sizeof(struct in6_addr) },
2747 [RTA_PREFSRC] = { .len = sizeof(struct in6_addr) },
2748 [RTA_OIF] = { .type = NLA_U32 },
2749 [RTA_IIF] = { .type = NLA_U32 },
2750 [RTA_PRIORITY] = { .type = NLA_U32 },
2751 [RTA_METRICS] = { .type = NLA_NESTED },
2752 [RTA_MULTIPATH] = { .len = sizeof(struct rtnexthop) },
2753 [RTA_PREF] = { .type = NLA_U8 },
2754 [RTA_ENCAP_TYPE] = { .type = NLA_U16 },
2755 [RTA_ENCAP] = { .type = NLA_NESTED },
2756 [RTA_TABLE] = { .type = NLA_U32 },
2759 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2760 struct fib6_config *cfg)
2763 struct nlattr *tb[RTA_MAX+1];
2767 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2772 rtm = nlmsg_data(nlh);
2773 memset(cfg, 0, sizeof(*cfg));
2775 cfg->fc_table = rtm->rtm_table;
2776 cfg->fc_dst_len = rtm->rtm_dst_len;
2777 cfg->fc_src_len = rtm->rtm_src_len;
2778 cfg->fc_flags = RTF_UP;
2779 cfg->fc_protocol = rtm->rtm_protocol;
2780 cfg->fc_type = rtm->rtm_type;
2782 if (rtm->rtm_type == RTN_UNREACHABLE ||
2783 rtm->rtm_type == RTN_BLACKHOLE ||
2784 rtm->rtm_type == RTN_PROHIBIT ||
2785 rtm->rtm_type == RTN_THROW)
2786 cfg->fc_flags |= RTF_REJECT;
2788 if (rtm->rtm_type == RTN_LOCAL)
2789 cfg->fc_flags |= RTF_LOCAL;
2791 if (rtm->rtm_flags & RTM_F_CLONED)
2792 cfg->fc_flags |= RTF_CACHE;
2794 cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2795 cfg->fc_nlinfo.nlh = nlh;
2796 cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2798 if (tb[RTA_GATEWAY]) {
2799 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2800 cfg->fc_flags |= RTF_GATEWAY;
2804 int plen = (rtm->rtm_dst_len + 7) >> 3;
2806 if (nla_len(tb[RTA_DST]) < plen)
2809 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2813 int plen = (rtm->rtm_src_len + 7) >> 3;
2815 if (nla_len(tb[RTA_SRC]) < plen)
2818 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2821 if (tb[RTA_PREFSRC])
2822 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2825 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2827 if (tb[RTA_PRIORITY])
2828 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2830 if (tb[RTA_METRICS]) {
2831 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2832 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2836 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2838 if (tb[RTA_MULTIPATH]) {
2839 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2840 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2844 pref = nla_get_u8(tb[RTA_PREF]);
2845 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2846 pref != ICMPV6_ROUTER_PREF_HIGH)
2847 pref = ICMPV6_ROUTER_PREF_MEDIUM;
2848 cfg->fc_flags |= RTF_PREF(pref);
2852 cfg->fc_encap = tb[RTA_ENCAP];
2854 if (tb[RTA_ENCAP_TYPE])
2855 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2863 struct rt6_info *rt6_info;
2864 struct fib6_config r_cfg;
2865 struct mx6_config mxc;
2866 struct list_head next;
2869 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2873 list_for_each_entry(nh, rt6_nh_list, next) {
2874 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2875 &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2876 nh->r_cfg.fc_ifindex);
2880 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2881 struct rt6_info *rt, struct fib6_config *r_cfg)
2886 list_for_each_entry(nh, rt6_nh_list, next) {
2887 /* check if rt6_info already exists */
2888 if (rt6_duplicate_nexthop(nh->rt6_info, rt))
2892 nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2896 err = ip6_convert_metrics(&nh->mxc, r_cfg);
2901 memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2902 list_add_tail(&nh->next, rt6_nh_list);
2907 static int ip6_route_multipath_add(struct fib6_config *cfg)
2909 struct fib6_config r_cfg;
2910 struct rtnexthop *rtnh;
2911 struct rt6_info *rt;
2912 struct rt6_nh *err_nh;
2913 struct rt6_nh *nh, *nh_safe;
2918 int replace = (cfg->fc_nlinfo.nlh &&
2919 (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2920 LIST_HEAD(rt6_nh_list);
2922 remaining = cfg->fc_mp_len;
2923 rtnh = (struct rtnexthop *)cfg->fc_mp;
2925 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2926 * rt6_info structs per nexthop
2928 while (rtnh_ok(rtnh, remaining)) {
2929 memcpy(&r_cfg, cfg, sizeof(*cfg));
2930 if (rtnh->rtnh_ifindex)
2931 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2933 attrlen = rtnh_attrlen(rtnh);
2935 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2937 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2939 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2940 r_cfg.fc_flags |= RTF_GATEWAY;
2942 r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2943 nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2945 r_cfg.fc_encap_type = nla_get_u16(nla);
2948 rt = ip6_route_info_create(&r_cfg);
2955 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2961 rtnh = rtnh_next(rtnh, &remaining);
2965 list_for_each_entry(nh, &rt6_nh_list, next) {
2966 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2967 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2968 nh->rt6_info = NULL;
2971 ip6_print_replace_route_err(&rt6_nh_list);
2976 /* Because each route is added like a single route we remove
2977 * these flags after the first nexthop: if there is a collision,
2978 * we have already failed to add the first nexthop:
2979 * fib6_add_rt2node() has rejected it; when replacing, old
2980 * nexthops have been replaced by first new, the rest should
2983 if (cfg->fc_nlinfo.nlh) {
2984 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2986 cfg->fc_nlinfo.nlh->nlmsg_flags |= NLM_F_CREATE;
2994 /* Delete routes that were already added */
2995 list_for_each_entry(nh, &rt6_nh_list, next) {
2998 ip6_route_del(&nh->r_cfg);
3002 list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3004 dst_free(&nh->rt6_info->dst);
3006 list_del(&nh->next);
3013 static int ip6_route_multipath_del(struct fib6_config *cfg)
3015 struct fib6_config r_cfg;
3016 struct rtnexthop *rtnh;
3019 int err = 1, last_err = 0;
3021 remaining = cfg->fc_mp_len;
3022 rtnh = (struct rtnexthop *)cfg->fc_mp;
3024 /* Parse a Multipath Entry */
3025 while (rtnh_ok(rtnh, remaining)) {
3026 memcpy(&r_cfg, cfg, sizeof(*cfg));
3027 if (rtnh->rtnh_ifindex)
3028 r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3030 attrlen = rtnh_attrlen(rtnh);
3032 struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3034 nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3036 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3037 r_cfg.fc_flags |= RTF_GATEWAY;
3040 err = ip6_route_del(&r_cfg);
3044 rtnh = rtnh_next(rtnh, &remaining);
3050 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3052 struct fib6_config cfg;
3055 err = rtm_to_fib6_config(skb, nlh, &cfg);
3060 return ip6_route_multipath_del(&cfg);
3062 return ip6_route_del(&cfg);
3065 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3067 struct fib6_config cfg;
3070 err = rtm_to_fib6_config(skb, nlh, &cfg);
3075 return ip6_route_multipath_add(&cfg);
3077 return ip6_route_add(&cfg);
3080 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3082 return NLMSG_ALIGN(sizeof(struct rtmsg))
3083 + nla_total_size(16) /* RTA_SRC */
3084 + nla_total_size(16) /* RTA_DST */
3085 + nla_total_size(16) /* RTA_GATEWAY */
3086 + nla_total_size(16) /* RTA_PREFSRC */
3087 + nla_total_size(4) /* RTA_TABLE */
3088 + nla_total_size(4) /* RTA_IIF */
3089 + nla_total_size(4) /* RTA_OIF */
3090 + nla_total_size(4) /* RTA_PRIORITY */
3091 + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3092 + nla_total_size(sizeof(struct rta_cacheinfo))
3093 + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3094 + nla_total_size(1) /* RTA_PREF */
3095 + lwtunnel_get_encap_size(rt->dst.lwtstate);
3098 static int rt6_fill_node(struct net *net,
3099 struct sk_buff *skb, struct rt6_info *rt,
3100 struct in6_addr *dst, struct in6_addr *src,
3101 int iif, int type, u32 portid, u32 seq,
3102 int prefix, int nowait, unsigned int flags)
3104 u32 metrics[RTAX_MAX];
3106 struct nlmsghdr *nlh;
3110 if (prefix) { /* user wants prefix routes only */
3111 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3112 /* success since this is not a prefix route */
3117 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3121 rtm = nlmsg_data(nlh);
3122 rtm->rtm_family = AF_INET6;
3123 rtm->rtm_dst_len = rt->rt6i_dst.plen;
3124 rtm->rtm_src_len = rt->rt6i_src.plen;
3127 table = rt->rt6i_table->tb6_id;
3129 table = RT6_TABLE_UNSPEC;
3130 rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
3131 if (nla_put_u32(skb, RTA_TABLE, table))
3132 goto nla_put_failure;
3133 if (rt->rt6i_flags & RTF_REJECT) {
3134 switch (rt->dst.error) {
3136 rtm->rtm_type = RTN_BLACKHOLE;
3139 rtm->rtm_type = RTN_PROHIBIT;
3142 rtm->rtm_type = RTN_THROW;
3145 rtm->rtm_type = RTN_UNREACHABLE;
3149 else if (rt->rt6i_flags & RTF_LOCAL)
3150 rtm->rtm_type = RTN_LOCAL;
3151 else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3152 rtm->rtm_type = RTN_LOCAL;
3154 rtm->rtm_type = RTN_UNICAST;
3156 if (!netif_carrier_ok(rt->dst.dev)) {
3157 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3158 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3159 rtm->rtm_flags |= RTNH_F_DEAD;
3161 rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3162 rtm->rtm_protocol = rt->rt6i_protocol;
3163 if (rt->rt6i_flags & RTF_DYNAMIC)
3164 rtm->rtm_protocol = RTPROT_REDIRECT;
3165 else if (rt->rt6i_flags & RTF_ADDRCONF) {
3166 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3167 rtm->rtm_protocol = RTPROT_RA;
3169 rtm->rtm_protocol = RTPROT_KERNEL;
3172 if (rt->rt6i_flags & RTF_CACHE)
3173 rtm->rtm_flags |= RTM_F_CLONED;
3176 if (nla_put_in6_addr(skb, RTA_DST, dst))
3177 goto nla_put_failure;
3178 rtm->rtm_dst_len = 128;
3179 } else if (rtm->rtm_dst_len)
3180 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3181 goto nla_put_failure;
3182 #ifdef CONFIG_IPV6_SUBTREES
3184 if (nla_put_in6_addr(skb, RTA_SRC, src))
3185 goto nla_put_failure;
3186 rtm->rtm_src_len = 128;
3187 } else if (rtm->rtm_src_len &&
3188 nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3189 goto nla_put_failure;
3192 #ifdef CONFIG_IPV6_MROUTE
3193 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3194 int err = ip6mr_get_route(net, skb, rtm, nowait,
3201 goto nla_put_failure;
3203 if (err == -EMSGSIZE)
3204 goto nla_put_failure;
3209 if (nla_put_u32(skb, RTA_IIF, iif))
3210 goto nla_put_failure;
3212 struct in6_addr saddr_buf;
3213 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3214 nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3215 goto nla_put_failure;
3218 if (rt->rt6i_prefsrc.plen) {
3219 struct in6_addr saddr_buf;
3220 saddr_buf = rt->rt6i_prefsrc.addr;
3221 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3222 goto nla_put_failure;
3225 memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3227 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3228 if (rtnetlink_put_metrics(skb, metrics) < 0)
3229 goto nla_put_failure;
3231 if (rt->rt6i_flags & RTF_GATEWAY) {
3232 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3233 goto nla_put_failure;
3237 nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3238 goto nla_put_failure;
3239 if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3240 goto nla_put_failure;
3242 expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3244 if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3245 goto nla_put_failure;
3247 if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3248 goto nla_put_failure;
3250 if (lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3251 goto nla_put_failure;
3253 nlmsg_end(skb, nlh);
3257 nlmsg_cancel(skb, nlh);
3261 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3263 struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3266 if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3267 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3268 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3272 return rt6_fill_node(arg->net,
3273 arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3274 NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3275 prefix, 0, NLM_F_MULTI);
3278 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3280 struct net *net = sock_net(in_skb->sk);
3281 struct nlattr *tb[RTA_MAX+1];
3282 struct rt6_info *rt;
3283 struct sk_buff *skb;
3286 int err, iif = 0, oif = 0;
3288 err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3293 memset(&fl6, 0, sizeof(fl6));
3296 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3299 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3303 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3306 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3310 iif = nla_get_u32(tb[RTA_IIF]);
3313 oif = nla_get_u32(tb[RTA_OIF]);
3316 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3319 struct net_device *dev;
3322 dev = __dev_get_by_index(net, iif);
3328 fl6.flowi6_iif = iif;
3330 if (!ipv6_addr_any(&fl6.saddr))
3331 flags |= RT6_LOOKUP_F_HAS_SADDR;
3333 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3336 fl6.flowi6_oif = oif;
3338 if (netif_index_is_l3_master(net, oif)) {
3339 fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3340 FLOWI_FLAG_SKIP_NH_OIF;
3343 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3346 skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3353 /* Reserve room for dummy headers, this skb can pass
3354 through good chunk of routing engine.
3356 skb_reset_mac_header(skb);
3357 skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3359 skb_dst_set(skb, &rt->dst);
3361 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3362 RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3363 nlh->nlmsg_seq, 0, 0, 0);
3369 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3374 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3375 unsigned int nlm_flags)
3377 struct sk_buff *skb;
3378 struct net *net = info->nl_net;
3383 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3385 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3389 err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3390 event, info->portid, seq, 0, 0, nlm_flags);
3392 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3393 WARN_ON(err == -EMSGSIZE);
3397 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3398 info->nlh, gfp_any());
3402 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3405 static int ip6_route_dev_notify(struct notifier_block *this,
3406 unsigned long event, void *ptr)
3408 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3409 struct net *net = dev_net(dev);
3411 if (!(dev->flags & IFF_LOOPBACK))
3414 if (event == NETDEV_REGISTER) {
3415 net->ipv6.ip6_null_entry->dst.dev = dev;
3416 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3417 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3418 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3419 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3420 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3421 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3423 } else if (event == NETDEV_UNREGISTER &&
3424 dev->reg_state != NETREG_UNREGISTERED) {
3425 /* NETDEV_UNREGISTER could be fired for multiple times by
3426 * netdev_wait_allrefs(). Make sure we only call this once.
3428 in6_dev_put(net->ipv6.ip6_null_entry->rt6i_idev);
3429 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3430 in6_dev_put(net->ipv6.ip6_prohibit_entry->rt6i_idev);
3431 in6_dev_put(net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3442 #ifdef CONFIG_PROC_FS
3444 static const struct file_operations ipv6_route_proc_fops = {
3445 .owner = THIS_MODULE,
3446 .open = ipv6_route_open,
3448 .llseek = seq_lseek,
3449 .release = seq_release_net,
3452 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3454 struct net *net = (struct net *)seq->private;
3455 seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3456 net->ipv6.rt6_stats->fib_nodes,
3457 net->ipv6.rt6_stats->fib_route_nodes,
3458 net->ipv6.rt6_stats->fib_rt_alloc,
3459 net->ipv6.rt6_stats->fib_rt_entries,
3460 net->ipv6.rt6_stats->fib_rt_cache,
3461 dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3462 net->ipv6.rt6_stats->fib_discarded_routes);
3467 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3469 return single_open_net(inode, file, rt6_stats_seq_show);
3472 static const struct file_operations rt6_stats_seq_fops = {
3473 .owner = THIS_MODULE,
3474 .open = rt6_stats_seq_open,
3476 .llseek = seq_lseek,
3477 .release = single_release_net,
3479 #endif /* CONFIG_PROC_FS */
3481 #ifdef CONFIG_SYSCTL
3484 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3485 void __user *buffer, size_t *lenp, loff_t *ppos)
3492 net = (struct net *)ctl->extra1;
3493 delay = net->ipv6.sysctl.flush_delay;
3494 proc_dointvec(ctl, write, buffer, lenp, ppos);
3495 fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3499 struct ctl_table ipv6_route_table_template[] = {
3501 .procname = "flush",
3502 .data = &init_net.ipv6.sysctl.flush_delay,
3503 .maxlen = sizeof(int),
3505 .proc_handler = ipv6_sysctl_rtcache_flush
3508 .procname = "gc_thresh",
3509 .data = &ip6_dst_ops_template.gc_thresh,
3510 .maxlen = sizeof(int),
3512 .proc_handler = proc_dointvec,
3515 .procname = "max_size",
3516 .data = &init_net.ipv6.sysctl.ip6_rt_max_size,
3517 .maxlen = sizeof(int),
3519 .proc_handler = proc_dointvec,
3522 .procname = "gc_min_interval",
3523 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3524 .maxlen = sizeof(int),
3526 .proc_handler = proc_dointvec_jiffies,
3529 .procname = "gc_timeout",
3530 .data = &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3531 .maxlen = sizeof(int),
3533 .proc_handler = proc_dointvec_jiffies,
3536 .procname = "gc_interval",
3537 .data = &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3538 .maxlen = sizeof(int),
3540 .proc_handler = proc_dointvec_jiffies,
3543 .procname = "gc_elasticity",
3544 .data = &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3545 .maxlen = sizeof(int),
3547 .proc_handler = proc_dointvec,
3550 .procname = "mtu_expires",
3551 .data = &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3552 .maxlen = sizeof(int),
3554 .proc_handler = proc_dointvec_jiffies,
3557 .procname = "min_adv_mss",
3558 .data = &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3559 .maxlen = sizeof(int),
3561 .proc_handler = proc_dointvec,
3564 .procname = "gc_min_interval_ms",
3565 .data = &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3566 .maxlen = sizeof(int),
3568 .proc_handler = proc_dointvec_ms_jiffies,
3573 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3575 struct ctl_table *table;
3577 table = kmemdup(ipv6_route_table_template,
3578 sizeof(ipv6_route_table_template),
3582 table[0].data = &net->ipv6.sysctl.flush_delay;
3583 table[0].extra1 = net;
3584 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3585 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3586 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3587 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3588 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3589 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3590 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3591 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3592 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3594 /* Don't export sysctls to unprivileged users */
3595 if (net->user_ns != &init_user_ns)
3596 table[0].procname = NULL;
3603 static int __net_init ip6_route_net_init(struct net *net)
3607 memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3608 sizeof(net->ipv6.ip6_dst_ops));
3610 if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3611 goto out_ip6_dst_ops;
3613 net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3614 sizeof(*net->ipv6.ip6_null_entry),
3616 if (!net->ipv6.ip6_null_entry)
3617 goto out_ip6_dst_entries;
3618 net->ipv6.ip6_null_entry->dst.path =
3619 (struct dst_entry *)net->ipv6.ip6_null_entry;
3620 net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3621 dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3622 ip6_template_metrics, true);
3624 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3625 net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3626 sizeof(*net->ipv6.ip6_prohibit_entry),
3628 if (!net->ipv6.ip6_prohibit_entry)
3629 goto out_ip6_null_entry;
3630 net->ipv6.ip6_prohibit_entry->dst.path =
3631 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3632 net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3633 dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3634 ip6_template_metrics, true);
3636 net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3637 sizeof(*net->ipv6.ip6_blk_hole_entry),
3639 if (!net->ipv6.ip6_blk_hole_entry)
3640 goto out_ip6_prohibit_entry;
3641 net->ipv6.ip6_blk_hole_entry->dst.path =
3642 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3643 net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3644 dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3645 ip6_template_metrics, true);
3648 net->ipv6.sysctl.flush_delay = 0;
3649 net->ipv6.sysctl.ip6_rt_max_size = 4096;
3650 net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3651 net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3652 net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3653 net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3654 net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3655 net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3657 net->ipv6.ip6_rt_gc_expire = 30*HZ;
3663 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3664 out_ip6_prohibit_entry:
3665 kfree(net->ipv6.ip6_prohibit_entry);
3667 kfree(net->ipv6.ip6_null_entry);
3669 out_ip6_dst_entries:
3670 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3675 static void __net_exit ip6_route_net_exit(struct net *net)
3677 kfree(net->ipv6.ip6_null_entry);
3678 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3679 kfree(net->ipv6.ip6_prohibit_entry);
3680 kfree(net->ipv6.ip6_blk_hole_entry);
3682 dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3685 static int __net_init ip6_route_net_init_late(struct net *net)
3687 #ifdef CONFIG_PROC_FS
3688 proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3689 proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3694 static void __net_exit ip6_route_net_exit_late(struct net *net)
3696 #ifdef CONFIG_PROC_FS
3697 remove_proc_entry("ipv6_route", net->proc_net);
3698 remove_proc_entry("rt6_stats", net->proc_net);
3702 static struct pernet_operations ip6_route_net_ops = {
3703 .init = ip6_route_net_init,
3704 .exit = ip6_route_net_exit,
3707 static int __net_init ipv6_inetpeer_init(struct net *net)
3709 struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3713 inet_peer_base_init(bp);
3714 net->ipv6.peers = bp;
3718 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3720 struct inet_peer_base *bp = net->ipv6.peers;
3722 net->ipv6.peers = NULL;
3723 inetpeer_invalidate_tree(bp);
3727 static struct pernet_operations ipv6_inetpeer_ops = {
3728 .init = ipv6_inetpeer_init,
3729 .exit = ipv6_inetpeer_exit,
3732 static struct pernet_operations ip6_route_net_late_ops = {
3733 .init = ip6_route_net_init_late,
3734 .exit = ip6_route_net_exit_late,
3737 static struct notifier_block ip6_route_dev_notifier = {
3738 .notifier_call = ip6_route_dev_notify,
3739 .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
3742 void __init ip6_route_init_special_entries(void)
3744 /* Registering of the loopback is done before this portion of code,
3745 * the loopback reference in rt6_info will not be taken, do it
3746 * manually for init_net */
3747 init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3748 init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3749 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3750 init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3751 init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3752 init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3753 init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3757 int __init ip6_route_init(void)
3763 ip6_dst_ops_template.kmem_cachep =
3764 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3765 SLAB_HWCACHE_ALIGN, NULL);
3766 if (!ip6_dst_ops_template.kmem_cachep)
3769 ret = dst_entries_init(&ip6_dst_blackhole_ops);
3771 goto out_kmem_cache;
3773 ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3775 goto out_dst_entries;
3777 ret = register_pernet_subsys(&ip6_route_net_ops);
3779 goto out_register_inetpeer;
3781 ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3785 goto out_register_subsys;
3791 ret = fib6_rules_init();
3795 ret = register_pernet_subsys(&ip6_route_net_late_ops);
3797 goto fib6_rules_init;
3800 if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3801 __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3802 __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3803 goto out_register_late_subsys;
3805 ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3807 goto out_register_late_subsys;
3809 for_each_possible_cpu(cpu) {
3810 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3812 INIT_LIST_HEAD(&ul->head);
3813 spin_lock_init(&ul->lock);
3819 out_register_late_subsys:
3820 unregister_pernet_subsys(&ip6_route_net_late_ops);
3822 fib6_rules_cleanup();
3827 out_register_subsys:
3828 unregister_pernet_subsys(&ip6_route_net_ops);
3829 out_register_inetpeer:
3830 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3832 dst_entries_destroy(&ip6_dst_blackhole_ops);
3834 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3838 void ip6_route_cleanup(void)
3840 unregister_netdevice_notifier(&ip6_route_dev_notifier);
3841 unregister_pernet_subsys(&ip6_route_net_late_ops);
3842 fib6_rules_cleanup();
3845 unregister_pernet_subsys(&ipv6_inetpeer_ops);
3846 unregister_pernet_subsys(&ip6_route_net_ops);
3847 dst_entries_destroy(&ip6_dst_blackhole_ops);
3848 kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);